“like the children in the game of hot boiled beans and very good butter1, he was warm in his search when he saw the Tower, and cold when he didn’t see it”
- Charles Dicken, The Mystery of Edwin Drood

Two quick asks before we dive deeper:

1. I am in Philadelphia from 8-11th September. If you are a scientist/engineer/founder in the area or know someone who is, hit me up.

2. I am in San Francisco from 12th September onwards. Hit me up then if you are around.

******
At BioCompute, we use solid state nanopores to retrieve data from DNA.

Nanopores (i.e. nano-sized holes on a surface) typically use voltage as a physical property for moving charged molecules like DNA through it, so that fluctuations in current/voltage signals can be used to determine properties of the molecule.

Solid state nanopores (nanopores made in inorganic materials like silicon and graphene) while offering engineering advantages face a few different challenges, compared to biological nanopores. This recently published Nature Bioengineering paper2 recently talks about interesting ways to leverage temperature to address these challenges (Shoutout to Sajeer, first author on the paper, for sharing a copy with me!).

1. How does the molecule enter the pore?
   
   When a molecule passes through a pore, there has to be some physical parameter controlling how it enters the pore and in what conformation. In biological nanopores, this is addressed by placing a protein (like helicase) on the pore and using it to unzip DNA/unfold proteins, slowing down the movement of these molecules through the pore. However, functionalizing molecules on solid state nanopores is challenging.
   
   This paper explores how strategic temperature control could be used to unzip DNA/unfold proteins, without damaging the molecules.

   

   Figure 1: (a) Left - DNA being unzipped by increased temperature (b) Protein being unfolded by increased temperature in a solid state nanopore
   Source: https://www.nature.com/articles/s44222-026-00485-9
   

2. How long does it take for the molecule to pass through the pore?
   
   DNA (and other biomolecules) pass through nanopores fairly quickly (microseconds), faster than any sensing system can deduce meaningful signals from it.
   
   This paper recommends two temperature based approaches to slow down molecules passing through the pore:
   i) local cooling in the nanopores can increase dwell times (time that a molecule hangs around in the pore) making it possible to get more snapshots of the molecule
   ii) temperature gradients can cause DNA to expand and hence take longer to pass through the pore (similar to the expandomer approach that Roche’s new sequencer uses)
   

   

   Figure 2: Controlling the speed of translocation (i.e. the process of a molecule moving from one side of the pore to the other)
   Source: https://www.nature.com/articles/s44222-026-00485-9
   
   3. Capture Rate
   
   How do we get more of the same molecule to pass through the pore at a given time?
   
   DNA moves from warmer to colder regions in a solution. This behaviour can be leveraged to enable the accumulation of a large concentration of DNA near the pore, thus enabling better detection.

   

   Figure 3: Enhancing capture rate through thermal gradient
   Source: https://www.nature.com/articles/s44222-026-00485-9
   
   4. Pore Sizes
   
   Size (i.e. diameter) is an important parameter of any nanopore system. It determines what kinds of molecules it can detect, and at what resolution.

   

   Figure 4: Illustration of pore sizes; Source: Plenty of Room

   
   Pore sizes are usually only controllable during the pore fabrication process3 and it’s hard to get reliable sizes each time. The authors argue that temperature could be used to enlarge or shrink pore sizes, thus enabling size control at the device level post-fabrication.
   
   5. The million dollar SNR4 question
   
   Let’s first understand what kinds of noises typically show up in a solid state nanopore systems.
   
   i) Johnson Nyquist Noise (also called Thermal Noise)
   This noise arises from the intrinsic resistance of the electrolyte and the pore. It is frequency-independent (white noise) and sets a fundamental floor, proportional to temperature and inversely proportional to resistance (4kT/R). Lower pore/electrolyte resistance means lower thermal noise, but that trades off against signal amplitude.
   
   ii) 1/f Noise (also called Flicker Noise)
   This noise is attributed to surface charge fluctuations5 at the pore walls. It dominates below a crossover frequency that is typically in the 1–10 kHz range depending on the device; it’s especially problematic for slow measurements. Magnitude scales inversely with pore area and depends strongly on surface chemistry and passivation (coating on the surface of the membrane) quality.

   
   (iii) Dielectric Noise
   Nanopores are usually made on a membrane deposited on a support chip, for example silicon nitride membrane on a silicon chip. Dielectric noise comes from energy dissipation (loss) in the membrane/support chip material as they are not perfect insulators. It becomes significant at mid-to-high frequencies (i.e. 10 kHz to 90 kHz) and is sensitive to chip design.
   
   (iv) Amplifier/Instrumentation Noise
   
   An amplifier converts and boosts the tiny ionic current through the nanopore into a larger, measurable voltage signal. In nanopore setups, it also simultaneously applies and holds (”clamps”) the driving voltage across the membrane.
   
   The current amplifier contributes voltage and current noise from its input stage electronics, independent of the pore/chip. As this stage handles the signal when it's at its smallest, any noise generated here gets added directly to the raw signal at its weakest point, and then gets amplified right along with the real signal by every subsequent stage. Noise introduced later in the amplification chain matters far less, since by that point the real signal has already been boosted well above it. Amplifier noise sets a baseline noise floor even with an ideal (infinite resistance, zero capacitance) pore.

   
   (v) Capacitive Noise
   
   Any two conductive surfaces (i.e. surfaces that allow electrons to pass through) separated by an insulator form a capacitor. In the context of a nanopore, like the electrolyte on either side of the silicon membrane, separated by the insulating membrane itself. When you apply a voltage across them, charge accumulates on each side. Capacitance (C) is a measure of how much electric charge a system can store for a given voltage applied across it.
   
   Capacitive noise arises from the interaction between the amplifier’s input voltage noise and the total capacitance of the system. It scales with frequency squared, so dominates at high frequencies (100 kHz and above) and is the main bandwidth-limiting factor for fast single-molecule translocation measurements.
   
   (vi) Shot Noise
   Even at a perfectly steady, constant applied voltage, the exact number of ions passing through in any tiny time window fluctuates randomly. This randomness in the timing/count of individual charge-carrying events is what creates shot noise. It is usually negligible relative to other sources in typical nanopore conditions, though it can become relevant at very low ionic currents or very high bandwidths.
   
   If we use temperature as a way to control the parameters listed above, do we not get more noise?
   
   The authors suggest a few ways to get around this.
   
   i. Ultra low capacitance device architectures
   This includes reducing the size of the membrane (not the chip) on which the pore is made, thick insulating layers below and around the membrane, and fully insulating substrates (like glass) for the chip instead of silicon. Glass nanopores shows high SNR even at high temperatures.
   
   ii. Post acquisition low pass filtering
   A low-pass filter is a circuit that lets signals with a lower frequency than its cutoff pass through while reducing high-frequency signals.
   
   iv. Engineer thermal environments in flow cells where the solid-state nanopore is kept between chemically inert heat-transfer fluids
   
   v. Leverage on-chip, sub-micrometer thermometers6 that can measure nanoscale temperature variations and using it to construct a closed loop thermal feedback network that can be used to track pore destabilization due to thermal effects.
   
   *****
   This paper is largely theoretical. I would be keen to see how these approaches are tested experimentally. If you/someone in your network is testing these, I would love to chat.
   *****
   If you enjoyed reading this, two bonus articles to check out:
   1. Data storage in DNA origami
   2. DNA encryption inside cells

Subscribe now

My chapter focuses on end to end rewritable DNA data storage and the unit economics that makes it possible to build it today (vis a vis DNA synthesis), which is the fundamental thesis on which BioCompute is built. And it’s the first time in the world that such an analysis has been published, so check it out!

Over the next few weeks, we will be breaking down the concepts in my chapter in bite sized formats that is easily accessible.

*******

“They’re everything to each other for this short stretch of time because they’re all there is. They are each to the other a representative of the human race - they each have to suffice for billions of people”
- Samantha Harvey, Orbital

Harvey was writing about six astronauts from different parts of the world on a spaceship orbiting our planet. She could have easily been describing the Synbiobeta delegation that I visited China with last week.

It was an eye-opening trip for many reasons.

1. The collaborative ecosystem
   For the first time ever, I saw academic institutions, startups, startup enablers (like incubators and accelerators) and government work in such close coordination to ensure that businesses succeed. There are multiple reasons why this is happening, but highlighting the ones I found most relevant within the context of my interactions.
   
   i) The KPIs of local government officials are tied to metrics like getting more foreign entrepreneurs set up in their locality, accessing grants from the federal government for specific startup programs, i.e. they are evaluated more like a private sector employee than anywhere else in the world.

   

   
   ii) The state owns the means of production including land, this gives them an incentive to ensure that the value of their land grows exponentially by enabling businesses. This also gives government bodies a fair reason to build regulation in a way that enables entrepreneurs to succeed, for instance Kevin Zhao founder of Qi Biodesign told us that he was able to get export licenses for his genetically modified agricultural products by demonstrating the commercial value of his product (and not through crony capitalism), and now this regulation supports other companies in the same domain.

   
   iii) Academic institutions hire dedicated folks to enable tech transfer, instead of expecting an academic/an administrator to pick up this role. For instance, Cam Watson who joined us on this trip is working with XJTLU to enable their students and faculty to start and run their own deep tech companies.
   
   Breakneck by Dan Wang would be a good starting point if you want to dive deeper into understanding China’s growth and ecosystem.

2. Openness to do business with the rest of the world
   There is a perception in the rest of the world about China that they are closed to outsiders and don’t want to engage/do businesses with us. I learnt that nothing could be farther from the truth than that.

   
   The hospitality I experienced in China, from individuals as well as from the organizations we visited is impeccable (at one of the dinners, when I said out loud that I really loved their tofu skin, they got an extra bowl of that made for me xD). Moreover, many of the government officials I spoke to were eager to help us access government grants in China; the incubators we visited wanted to enable setting up facilities for foreign companies; the organizations we visited were actively investing in international business partnerships.
   
   It’s a little tricky to navigate the social landscape if you don’t speak Mandarin and are not familiar with the cultural nuances, but that’s operationally solvable if you know what you want to get done.

3. Unit economics advantage
   China has a clear cost arbitrage when it comes to scaling up. We have seen this in microelectronics manufacturing and assembly before. And we are now seeing it in biomanufacturing. They produce fermented products at a 1000th of the cost anywhere else in the world, while adhering to all the safety standards.

   
   This is partly due to the availability of cheap labour, but it’s also because they have innovated on manufacturing processes and hardware, and they built an ecosystem that enables them to iterate real quick when a customer requirement comes in.
   
   We are in early conversations with some Chinese companies for partnerships, more on that soon!
   

4. The Great Wall of China is really real
   There are wonders of the world that you hear about, see in photos and videos and they attain a certain unreal quality in your mind. But when you see it with your own eyes, your belief in human ingenuity grows a little bit.

   

There is something life-altering about sitting at a dinner table with folks from wildly different parts of the world, across ages, backgrounds, and lived experiences starting to talk about biotech and gently slipping into talking about our biggest dreams, fears, frustrations and what makes us human across borders. Creating spaces like these to come together is incredible given the heat of the geopolitical tensions, and I am grateful John Cumbers made it happen (like he always does).

*A huge shoutout to Lynn and Emily from Globaltech IP for being our generous hosts, if you are looking to plan a trip to China let them be your sherpas. Also, make sure you have your VPNs, Alipay and WeChat set up before you land in China (that will make your trip 100x more convenient).

******
We have wrapped up our Bangalore operations, and I am taking a break at home (okay, as much of a break a founder can take xD) before heading out to the US. And we are building V1 of our storage device in the mean time.

Excited for the road ahead!

Subscribe now

“Beauty is terror. Whatever we call beautiful, we quiver before it”
- Donna Tart, The Secret History

In 2024, I had a crazy dream. Make data storage and compute infrastructure as space and energy efficient as biological systems are. Step 1 being DNA data storage, and the final frontier being biological qubits at room temperature.

That made the 22 year old me quiver with excitement, nervousness and some (steely) resolve. BioCompute was born.

Over the last two years, I put together an incredible team, raised money, set up our own lab, ran thousands of iterations, built a prototype, partnered with great collaborators, and proved that we can encode data into DNA at costs that are several orders of magnitude lower than the leading players in the space, and that customers are excited about what we are building (and want it yesterday).

We are the first lab in India to go after the audacious problem of DNA data storage, and have leveraged the resources in India over the last 2 years to get to an end to end prototype.

And now it’s time for a step change.

As we start assembling the V1 of BioCompute chips, we are relocating to San Francisco to leverage the capital, talent and customer risk appetite of the world’s startup capital. Here is a quick video that Vyom made to tell our story.

While India is now starting to invest in deep tech through RDI fund and the likes, we don’t think the ecosystem is ready for a product like ours. India often likes to play safe, looking at what has been built in the West and adapting it to our socio-cultural and economic landscape. But what we need to build a new age data storage hardware company, and take on Goliaths like IBM, is an ecosystem that is built on abundance and takes high-risk-high-reward bets.

That unfortunately means having to let go of our most valuable asset, the team here in Bangalore. These are folks who have spent most of their waking hours thinking, building, troubleshooting and most importantly enjoying the challenges that come with having a once in a lifetime opportunity to shape data infrastructure.

Their exit interviews were perhaps the hardest conversations of my life. I needed to call up three of my best friends for reassurance the night before I let the cat out of the bag.

And yet, the strongest signal that we have great company culture is that all of these conversations (as shocking and sad as they were) were reflective, generative and most importantly kind, compassionate and optimistic (about the future of the world, and our individual futures). All of them ended on ‘we would like to work together again someday’.

Some of these folks are now looking for their next thing. I am happy to introduce you to them if you are looking to hire.

It’s an exciting (and nerve-wrecking) chapter ahead as I head out to San Francisco and build on the product development we have done in the last few years, and take our first chip to customers. If you made the move recently, I would like to chat about all things relocation.

********
Some of my learnings from year 2 of company building:

1. Get a competent ops and partnerships person very early on, it’s exhausting for the founder to be involved in everyday ops

2. Fundraising doesn’t depend on tech progress alone, it’s a game of storytelling, marketing, and (many a times) FOMO.

3. FOMO can only be generated if relationships are built and nurtured. Meet people where they are, try to become friends with them, share updates regularly

4. Founder health == company health. Getting a health coach and being consistent with workouts (even during travel) was the highest ROI thing I started last year.

5. People, regardless of how high agency, require a sense of direction; and want the founder to put their foot down and make a decision in times of uncertainty. Decision making authority needs to be clearly emphasized.

6. Kindness is an underrated superpower.

What I am looking to get better at in the next year of company building:
1. Tell the most compelling stories about the vision of the company

2. Be more consistent with communicating with external stakeholders

3. Spend more time on hiring, fundraising, general direction setting and outsource more repeatable tasks

4. Trust my intuition a lot more when it comes to people, processes and decisions

If you have resources on these/some takeaways from experience, hit me up. I am all ears.

Subscribe now

So what have we been up to at BioCompute between Feb 2026 and now?

(a) Product
i. We built an end to end working prototype at the lowest cost for DNA data storage in the world. This is CRAZY exciting. The next step is to bring this prototype onto a chip.

ii. Tried to build an electronic lab notebook (ELN) as a separate product. The rationale was that we are building ELNs for our internal use anyways, why not commercialize it and use it as a steady revenue stream as well as a gateway into pharma? We spent ~2 months on it and then scrapped it because it became a distraction from our main product.

(b) Customers
i. Signed on two media production studios for early pilots (this came about when I spent two months in the US); figured out that media archival is our niche starting point as we head towards data centers

ii. Pitched to a large data infra player in Saudi, they were excited about working with us, but they pressed pause because of the conflict

(c) Team/Hiring
i. Hired a web developer who ended up growing into a software developer in the company. Abhishek ended up building our new website as well. Check it out here.

ii. Brought in an intern from National Technological University (NTU) Singapore, she helped automate many of our lab workflows.

iii. Tried to hire a senior scientist from a large pharma company. While she was excited about the role and was a great fit, she backed out due to family pressure. This showed me that hiring talent in India is not just about convincing the candidate.

iv. Brought in consultants for wet lab and marketing/storytelling. While the consultants were great people, I severely underestimated the time it would take for them to build context and then contribute to our work.

v. Spent time with a potential CTO candidate, and figured out it’s not a good fit; I learnt that there are some things you just don’t catch up about a person (in spite of reference checks, conversations with the team and advisors etc.) unless you personally spend a few days together

vi. Hired an operations associate and unfortunately had to revoke the offer

(d) Partnerships

i. Signed MOUs with IISc (Prof. Manoj Varma) and IIT Delhi (Prof. Ishaan Gupta) to accelerate product development by leveraging their research infrastructure. While the labs are great at what they do, I underestimated how much we would have to work on simple things like coordination, making sure people show up on time, clarity of deliverables (which is very different in academic settings vs a startup). That led to a bunch of hiccups, but we managed to get interesting stuff done.

ii. Narayan (R&D head Ather) joined us as a product advisor

(e) Grants, Accelerators, Competitions

i. I was selected as a Next Gen Bio Leader for the Synbiobeta Conference (the world’s largest synthetic bio conference) this year, and had the opportunity to present the work we do at BioCompute and build some fun relationships

ii. Won the People’s Choice Award at the BITSync Startup Pitch (the largest gathering of BITSians in the US)

iii. Wrapped up the Nidhi Prayas Grant with CCAMP

iv. Got accepted into the Rebalance accelerator and pitched at their demo day recently

v. Franci and Naveen represented BioCompute at the final round of Bharat Innovates in Mumbai

vi. Got selected for the second round of Karnataka ELEVATE Grant

vii. Found out that BIRAC is pausing BIG Grants after we spent a decent amount of time submitting an application

What’s next? More on that soon.

Subscribe now

Cloud lives on servers, which are storage and compute devices mounted on racks, in data centers. This podcast series by Google gives you a sense of what a data center feels like.

While the idea of the cloud was floated in the 1990s, it began capturing the market in the post .com-bubble era, where servers put together in the warehouses of Amazon and Google began to first power the internal operations of these companies, and then in 2006 emerged as a service.

Just like how the steam engines ushered in Industrial Revolution 1.0 producing a shift from manual agriculture to mechanized crop production; the widespread access to electricity brought in Industrial Revolution 2.0 of mass manufacturing and consumption; and access to the internet and personal computers brought in the Industrial Revolution 3.0 of democratized access to compute; the emergence and widespread use of AI enabled by the shift of enterprises from in-house server infrastructure towards cloud is the biggest industrial transformation of the 21st century so far.

Apart from AI applications, the cloud has enabled the development of streaming services (Netflix etc), self-driving cars (Waymo etc.) and accelerated drug discovery for pharma companies.

What the cloud has effectively done is to take the operational expenses out of the hands of individual companies and concentrate them in data centers located anywhere in the world. Over the last few years, investments in data centers have skyrocketed, with $420 billion being invested in capex alone by the largest data center providers in 2025 alone, the number expected to reach $600 billion by end of 2026 ($80 billion already invested by Q1) and $7 Trillion by 2030.

Capex is only one part of the story. Opex is the other. While all the initial cloud data centers were built in the United States, companies slowly started branching out to Southeast Asia, India, Middle East and now Africa. While this is in part enabled by the governments in these regions offering generous special economic zone provisions to cloud companies and their co-location providers, this shift is primarily led by space, energy and water constraints in the United States.

Let’s look at each of these components one by one.

i. Space
How much space does a data center take up?
It ranges between 100,000 - 10,000,000 sq. ft depending on the scale of the data center. For context, a football field (not the American football xD) is 81,000 sq.ft. So a super large data center would be worth slightly more than 123 football fields.

The total land area on earth is 1.6 * 10^15 sq. ft. That feels like a lot and capable of accommodating all the data centers we would ever need, but not yet.

10% of this is taken up by icebergs (it’s obvious why data centers cannot be built on icebergs); 36% of this is agricultural land which is important for food security; 26% is forest land; 19% deserts; 8% grassland; 1% fresh water bodies and 1% human settlement.

Data centers started out by being built on grassland, but has now expanded to encroach on agricultural land, in addition to human settlements. Grassland and forests are great carbon sinks, and the loss of these implies rising global warming. Data centers are being built extensively on desert land (in Saudi Arabia, UAE etc) but there are real cooling challenges in deserts.

ii. Energy & Water

The total global data center energy consumption (as of 2025) is ~485 TWh, which can power, equivalent to what 139 million homes would consume in a year. When we break it down, the average data center consumes energy equivalent to that of 100,000 homes. Note that the biggest source of energy consumption and heat generation in data centers is the IT infrastructure (~60% of total energy), referring to the compute and storage devices, the server racks they are placed in and how they connect with each other.

The energy efficiency of a data center is typically measured in Power Usage Effectiveness (PUE), which is the ratio of total facility power by IT equipment power. PUE of 1 is ideal.

While conventional data centers (enterprise, smaller colocation) have very low power density (~5–15 kW/rack) and mediocre PUE (1.54–1.92), while AI- first data centers have soaring density (130–600 kW/rack) and better PUE (1.10–1.15 for hyperscale). This means that as you go up in size and AI-focus, energy consumption per rack explodes even as efficiency per unit of compute improves.

As a result of the PUE not being a perfect 1, waste heat is generated. This leads to the production of local heat islands (raising temperatures between 2-9 °C) in regions up to a 10 km radius from the data center. There are efforts to utilize waste heat from data centers and return them to the grid in some form, but it is yet to be implemented at scale, partly because of the infrastructure required to do that.

Data centers are typically cooled using water. For context, a Meta data center that opened in 2018 using 500,000 gallons of water per day (i.e.10% of the entire county’s water consumption) and all of Google’s data centers combined using 5 billion gallons of water across all its data centers in 2023, with 31 percent of its freshwater withdrawals coming from watersheds with medium or high water scarcity.

Out of the total water derived from different sources for data center cooling, at least 70% is consumed by evaporation and only the rest used in a closed loop. Nearly 57% of water used for cooling is derived from fresh water sources, with more recent efforts to leverage waste water. However, waste water typically has dissolved solids, salts and nutrients that can corrode data center equipment, necessitating water treatment before use, which again consumes energy in a different form. That is why fresh water is still the go-to source for cooling. Water Usage Effectiveness (WUE) is the analogue to PUE for water efficiency of data centers, and the average WUE achieved is 1.8 L/kWh.

While air cooling (i.e. using fans and Computer Room Air Conditioners to circulate cold air throughout a facility, removing heat generated by IT equipment) is an alternative, it is typically less energy efficient than water cooling, leading to it not being widely adopted.

iii. Hardware replacement

While typical data center hardware has reasonably long lifespans (GPUs up to 8 years, SSDs up to 10 years, HDDs up to 7 years and tapes for up to 30 years), they are replaced within an 18-24 month timeframe. This is partially due to reader obsolescence, improved speeds and pre-emption of degradation. While some of the hyperscalers are talking about extending the lifespan of their hardware, this then results in more water required for cooling.

E-waste associated with the use of LLMs alone is predicted to reach 5 million tons by 2030 (1.2 million if LLMs are used sparingly).

Note: The extended climate impact of space, water and energy consumption coupled with e-waste boils down to 180 million tons of CO2 emissions annually, roughly double that of the emissions of the aviation industry. While CO2 capture and carbon credits is being used a way to mitigate these emissions, these initiatives are only at 50 million tons capacity per year as of 2026.

iv. Wild Card Entry - Security

Given the large space footprint of data centers, they are quite conspicuous. While data centers have a six to seven layer security mechanism for entry into their premises and to prevent fire incidents, they are quite vulnerable to enemy attacks, as indicated by the the recent attack on AWS data centers in the Middle East.

Blast resistant architecture, surveillance and defense facilities similar to those used in military bases and decentralization are emerging as ways to counter this threat. Decentralization is quite hard to achieve with the current infrastructure capabilities, as there is clear operational advantages to concentrating IT infrastructure in specific locations (this is the premise for the widespread adoption of cloud as we discussed earlier).

First Principles Approaches to Addressing These Challenges

i. Build more water and energy efficient data centers/data center networks

The starting point is to measure how much energy and water is being consumed and for what purpose. This then enables data centers to cut down on wasteful consumption. This has led to the rise of many IoT sensor companies and their products being adopted by hyperscalers.

Once this is established, data centers can then go on to retro fit different technology assets, such as heat pumps to turn waste heat into usable energy, immersion cooling instead of evaporative cooling and direct to chip (D2C) cooling instead of rack level cooling. Startups like Uravu Labs are helping data centers get fresh water from air to reduce reliance on local fresh water sources.

In terms of data center construction, water circularity (in-house treatment plants), material sustainability, picking location that enable providers to spend lesser on cooling due to natural factors (such as the Khazna data center in Mazdar City, Abu Dhabi) are being adopted as pathways to greater energy and water efficiency.

An extension of these approaches is building flexible data center-power grid networks. This means that power is directed towards specific data centers based on their compute loads at a given point in time, preventing idle servers from eating up energy by operating at their fullest capacity. Companies like Emerald AI, led by Varun Sivaram who previously worked with the US Department of State for Energy and Innovation, are enabling this systemic transformation.

ii. Cheaper and infinitely scalable sources of energy

There is a certain physical limit to which retrofitting, grid networks and better construction can enable efficiency. So we then start thinking about sources of energy that are not as limiting as fossil fuels in terms of availability. Enter renewable energy sources.

Solar, wind and hydroelectric power have already been widely adopted by data center providers,. Geothermal and nuclear energy are being looked at as the next frontiers powering data centers,. Both the public and private sectors are investing heavily in nuclear fusion.

Batteries for long term energy storage are being looked at as an important infrastructure asset to enable the benefits of renewable energy to accrue for data center use.

iii. Build data centers in a place where space and energy is abundant

Even if energy is infinitely scalable and cheap, what do we do about cooling and the large tracts of terrestrial space that data centers take up?

Two things you could - build data centers under the ocean or in space.

Microsoft attempted the former between 2018 and 2020 but data centers on the ocean floor are challenging due to the impact that it can have on marine life, the coastal communities that rely on them for livelihoods and how it would change the ocean’s ability to act as a carbon sink. In addition, we barely understand our oceans in comparison to outer space, and hence a lot of the infrastructure needed to enable ocean floor data centers needs to be built from scratch.

However, the space narrative is different. The (nearly) unlimited access to solar energy, radiation for cooling (instead of convection) and almost infinite physical space for expansion, coupled with recent advancements in rocket technology and space manufacturing makes orbital data centers a compelling alternative. As this also directly contributes to inter-planetary civilizations, like the moon/mars settlements that SpaceX attempts to build, orbital data centers are being heavily invested in.

There are definitely challenges to implementation there. Cooling by radiation sounds compelling but is not as simple, it requires massive infrastructure to capture the energy and radiate it back. While SpaceForge has shown that plasma, that is super important for semiconductor manufacturing in space, can be generated without human intervention, there is a long way to go towards building fabs in space given their operational complexity. Data centers require regular maintenance (unless a very new kind of space data center hardware gets developed) and we need to figure out how to do it economically. In addition, data center hardware is not built for space environments due to poor radiation tolerance. While NVIDIA is attempting to build more radiation resistant GPUs for space environments, Jenson himself admits that the cost curves will take time to fall exponentially.

These are the approaches that the industry primarily looks at today.

But there is a fourth approach that goes deep into the root of the space, energy, water and security problems, all at once. It is the hardware that goes into data centers. Why don’t we tackle the problem at its roots by redesigning storage and compute hardware?

More of that in part 2 of this series

Subscribe now

References/Footnotes

1. Now, there is greater focus on data sovereignty, which means that cloud providers have to make sure that data of a particular country stays within the country

2. https://www.delloro.com/news/data-center-capex-surges-57-percent-in-2025-as-ai-deployments-accelerate/#:~:text=Data%20Center%20Capex%20Surges%2057,Oro%20Group%20%2D%20Dell’Oro%20Group

3. https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-7-trillion-dollar-data-center-build-out-how-industrials-can-capture-their-share

4. This is specifically with respect to cloud data centers. Sify built data centers in India as far back as 2000, initially as a way to promote consumer internet use but then to provide co-location services to companies like Wipro and Makemytrip. Now Sify is a co-location partner for some of the hyper-scalers.

5. https://programs.com/resources/data-center-statistics/

6. https://www.iea.org/reports/key-questions-on-energy-and-ai

7. https://www.congress.gov/crs-product/R48646#fn16

8. https://www.frontiersin.org/journals/sustainability/articles/10.3389/frsus.2022.1008583/full

9. https://www.researchgate.net/publication/403073048_The_data_heat_island_effect_quantifying_the_impact_of_AI_data_centers_in_a_warming_world

10. https://local.microsoft.com/wp-content/uploads/2022/06/Azure_HeatReUse_Infographic.pdf

11. https://www.lincolninst.edu/publications/land-lines-magazine/articles/land-water-impacts-data-centers/

12. https://arxiv.org/pdf/2304.03271

13. https://www.nature.com/articles/s41545-021-00101-w

14. https://blog.google/company-news/outreach-and-initiatives/sustainability/our-commitment-to-climate-conscious-data-center-cooling/

15. https://www.mckinsey.com/capabilities/operations/our-insights/operations-blog/keeping-cool-in-the-data-age

16. https://www.fwpcoa.org/content.aspx?page_id=5&club_id=859275&item_id=130961

17. https://www.delltechnologies.com/asset/en-in/products/servers/industry-market/idc-server-refresh-white-paper.pdf

18. https://www.nature.com/articles/s43588-024-00712-6

19. https://about.bnef.com/insights/finance/global-carbon-capture-capacity-due-to-rise-sixfold-by-2030/#:~:text=Today%2C%20most%20capture%20capacity%20is,CO2%20to%20meet%20their%20goals.

20. https://www.cnbc.com/2026/04/07/aws-iran-threats-us-tech-data-centers.html

21. https://silentinfotech.com/blog/data-center-7/role-of-iot-in-real-time-data-center-monitoring-238

22. https://international.eco.de/topics/datacenter/white-paper-utilization-of-waste-heat-in-the-data-center/

23. https://www.infosys.com/iki/perspectives/data-centers-look-immersion-cooling.html

24. https://attom.tech/direct-to-chip-liquid-cooling-how-it-works-and-application/#:~:text=Headroom%20for%20Future%20Density%20Growth%20As%20chip,higher%20compute%20density%20within%20existing%20physical%20space.

25. https://www.iea.org/reports/energy-and-ai/energy-supply-for-ai

26. https://news.microsoft.com/source/features/sustainability/6-projects-that-helped-microsoft-meet-its-renewable-energy-goal/

27. https://www.weforum.org/stories/2026/04/nuclear-energy-crisis-strait-of-hormuz-wef-small-modular-reactor/

28. https://about.fb.com/news/2024/08/new-geothermal-energy-project-to-support-our-data-centers/

29. https://www.iaea.org/newscenter/news/fusion-energy-in-2025-six-global-trends-to-watch

30. https://energy.mit.edu/wp-content/uploads/2024/09/MITEI_FusionReport_091124_final_COMPLETE-REPORT_fordistribution.pdf

31. https://www.datacenterknowledge.com/energy-power-supply/potential-energy-is-bess-the-answer-to-data-centers-gridlocked-future-

32. https://news.microsoft.com/source/features/sustainability/project-natick-underwater-datacenter/#:~:text=Microsoft’s%20Project%20Natick%20deployed%20an%20underwater%20datacenter,was%20tested%20and%20monitored%20for%20two%20years.

33. https://www.scientificamerican.com/article/data-centers-in-space/

34. https://www.reuters.com/business/aerospace-defense/spacexs-orbital-data-centers-could-face-same-hurdles-microsofts-abandoned-2026-04-01/

35. https://www.scientificamerican.com/article/the-push-to-make-semiconductors-in-space-just-took-a-serious-leap-forward/#:~:text=Space%20manufacturing%20is%20still%20in%20its%20early,reaches%20temperatures%20of%20around%201%2C832%20degrees%20Fahrenheit.

    https://www.pcmag.com/news/nvidia-debuts-an-ai-chip-for-space-based-data-centers-gtc-2026#:~:text=Select%20your%20favorite%20topics:,of%20Use%20and%20Privacy%20Policy.&text=In%20contrast%2C%20Nvidia’s%20CEO%20Jensen,been%20developing%20orbiting%20data%20centers.

1. We got our first paying customer.

Delivering the service to the customer was a crazy learning curve for us. But this experience has changed my mind about leveraging a service based model for early revenue generation, while we continue to build the product.

The need to offer a service to a single customer was too distracting for us. This is perhaps because we don’t have separate teams for R&D and services, perhaps because focusing on providing services doesn’t really tie into the big vision of building the world’s most efficient data storage hardware.

2. We hosted quite a lot of visitors.

Jacki, a biologist from Berkeley, flew down to spend a week with us in our lab. We learnt a lot from her about systematic approaches to doing science, what it takes to scale a proof of concept to a product. I hope to continue working with her moving ahead.

We collaborated with the School of Chemistry, Chemical Engineering and Biotechnology (CCEB) of Nanyang Technology University (NTU), Singapore as a part of their Innovation Ecosystem Program, to offer university students an exposure to startups in domains that are closely related to their field of study. They were super eager to learn about what we do, and how they can potentially initiate an R&D collaboration with us through some of their professors.

Image Description: Somehow we have Dr. Jacki, the NTU students and our team all in one frame

We have had quite a few scientists, engineers, ex-founders, operators (yes, I have now actually started using this term because it’s the shortest word for folks working on anything non-tech in the startup world xD) drop by, and I have enjoyed picking their brains on some of the challenges we have run into.

3. The Real Upheaval

Our -80 freezer had to be replaced (which has caused us a decent bit of delay on the wet lab front), and given that it doesn’t fit in the lift had to be hauled up the stairs. We are super convinced that when we build an office building of our own, we will create enough physical infrastructure for heavy equipment to be moved around.

Scientific endeavour is also built on the shoulders of a lot of these people who we don’t give enough credit to. It’s perhaps easier (and more privileged) to be a scientist than someone who has to lift lab equipment.

Shout out to Franci from our team for coordinating the effort, it’s not the easiest thing to manage.

4. Events
   a. I spoke at St. Joseph’s University in Bangalore on a panel at the Smart Chem 2026 conference. I shared what we are building at BioCompute and how folks from a bio/chem background can shape this industry. I then hung around with one of the attendees and chatted about how we can build a community bio lab in the city. Thanks, Praapti, for making this happen!
   
   b. We were shortlisted for Phase 2 of the Bharat Innovates Challenge being organized by the Ministry of Education India; I presented our work to a panel of academics and industry veterans at IIT Bombay on 22nd Jan.

   

   I also leveraged that opportunity to meet and chat with the iGEM team at IITB. It’s fascinating to see how they decide on their projects, plan experiments and execute them. I am all the more convinced that we need an iGEM equivalent in India.

   
   c. I was at the Sustainability Showcase event hosted by Biocon (one of the largest pharma companies in India) in collaboration with the Sustainability Mafia (the largest community of climate founders in India). I had an opportunity to chat about what we are building at BioCompute and discuss how Biocon can leverage sustainable data storage using DNA.

5. Some Team Stuff
   I realized that I haven’t introduced our most recent team members (who are not so recent anymore) to you on Substack. Here they are:
   
   i) Franci
   At BioCompute, Franci has been involved in wet lab experiments, automating repetitive manual processes in the lab, and streamlining our supply chain. I am constantly awed by her ability to learn new skills and implement feedback at an unimaginable pace.

   She holds an MSc in Human Genetics from the Central University of Punjab, and has previously studied gut microbes in Drosophila melanogaster (fancy word for fruit flies), worked on identifying the genetic basis of Lesch-Nyhan Syndrome at the Rajiv Gandhi Center for Biotechnology and conducted bioinformatic analysis of the shared genetic and epigenetic signatures in PCOS and Breast Cancer in specific north Indian populations. Outside work, Franci enjoys powerlifting and creating different forms of art.

   
   b) Pooja
   At BioCompute, Pooja has been working extensively on lab experiments to encode data in DNA and sequence it. She has an impeccable sense of observation, which is evident in the feedback she gives folks on the team (including me).

   She holds a B.Tech in Biotechnology from PES University, and has previously worked on the regulation of transcription in the process of photosynthesis in cyanobacterial species at the University of Hyderabad, and studied the virulence of three different mutants of the bacterium Salmonella Typhimurium in in vitro and in vivo contexts at the Indian Institute of Science (IISc). Outside work, Pooja enjoys reading, covering songs on her uke, graphite sketching and vibing to R&B music.
   

   We did a little Secret Santa in office this year (I have been told much to my chagrin that this is a very corporate practice), followed by a round of UNO and it was fun. It was a time to reflect on the culture we have built in the company so far, how much everyone values it (all the gifts were so thoughtful), and how we keep scaling this culture as we grow.

   
   We are hiring a Lead Bioengineer and talking to a bunch of interesting folks for the role, if you/someone you know would like to be part of a fun team working on audacious goals, do reach out to me for a chat.

   

   Image Description: The Secret Santa Gifts
   
   *******
   I spent the last three weeks wondering what I should be writing in the context of a new year (in terms of reflections and the likes) and couldn’t come up with anything but the generic stuff. So I have decided to skip that for now and laser-focus on scaling up our data storage.

   
   Exciting updates soon!

   
   

   Subscribe now

The last few weeks have been crazy.

1. I represented BioCompute at the MolecularArxiv Autumn School, and presented our first ever poster at an academic gathering (thanks, Jacki for helping out with the poster).

Image Description: Poster presentation, yes it was outdoors

Hosted in a hamlet in Corsica (an island in the South of France where Napoleon Bonaparte was born and raised), the autumn school brought together some of the smartest folks in DNA data storage for five days. Discussions and debates flowed against the backdrop of the scenic bay surrounding the Research Institute Scientists De Cargèse, a space started by particle physicist Maurice Levy to enable scientists to get away from labs and classrooms, and rediscover the joy of chatting about science in a friendly setting. Being locked in an island (with limited access to the internet, and unlimited access to starry skies) got everyone to shed the formality of traditional academic conferences and bond over fun stuff like teaching and learning languages in the community kitchen (Thanks Keshav for keeping me well fed with Indian food xD).

We are in talks to kick start a few collaborations with folks I met in Cargese, more on that soon!

I also spent a week in Paris meeting with academic labs at Sorbonne (that’s where our competitor Biomemory started), ESPCI and the Curie Institute.

2. Automation
We have automated our encoding pipeline, using our automated liquid handler and some scripts. This is a fancy way of saying you can enter any text data that you want to store, and you will have a robotic system programmed to figure out which edits to make to DNA and actually do them.

Now we are zooming into scale up mode. We are soon going to be the first folks to store TB scales of data in DNA (at reasonable unit economics), fingers crossed.

We are now designing accelerated aging studies to test how long the data stored in DNA can last without degradation (P.S. it’s a few hundred years but we are attempting to get to the exact number) so that we can enable many of you to store data that matters to you for your future generations.

Subscribe now

That’s why we are now expanding our blog to bring in folks who want to break down complex DNA and data storage concepts for the non-expert. This is meant to be an extension of some of the research heavy pieces that I have written for the blog earlier (you can check them out here, here and here).

I started thinking of this idea when I first spoke to Saraswathi, a masters student in biotechnology at the Indian Institute of Technology (IIT) Guwahati, in June this year. She asked me whether we are open to having folks from outside BioCompute write for our blog, and back then I was in doubt. This is something that I started almost as a personal blog, and grew into being a company blog. I was unsure whether folks from outside BioCompute would be able to bring the spirit of our work into the world with as much care as we do.

But Saraswathi kept at it relentlessly, writing multiple follow up emails to me, and offering to put together a draft essay without any guarantee that it would see the light of the day. And when I read her work, I knew this had to come into the world. So we worked together on editing it, and after 4ish months of back and forth, her piece on DNA synthesis (how it works with some illustrations) is here for you to read.

*******

Some definitions:

1. Nucleotide: Refers to A, T, G or C - fundamental units that the molecule DNA is made of

2. Codon: Refers to a set of three nucleotides in any order

The Building Blocks

Nucleotides are like Legos: they are building blocks used to build something of the creator’s fancy. Legos build castles, nucleotides build the very code that makes us all what we are—DNA. The entire blueprint for our existence is stored in these stretches of A, T, G, and C. So why can’t we try to store all the data we hoard and multiply, in the same format?

As the world relentlessly flies through this digital era, isn’t the most important thing accessible storage for all our data? DNA data storage is a booming industry owing to its immense potential in reducing the need for more silicon hardware. But one of the major struggles we face is the cost and chemical load that is brought in by the synthesis process of DNA.

how is DNA synthesized? The traditional methods involve the addition of free nucleotides to a growing sequence one at a time. This is done using 2 methods:
(a) chemical synthesis and (b) enzymatic synthesis.

Chemical Synthesis

In chemical synthesis, we have nucleotides that have one free end and one protected end. This is done to avoid random reactions between nucleotides. This protecting agent is removed at each step to allow the next nucleotide to attach to it. This is a highly accurate method owing to this protection-deprotection cycle.

Check out the image below for context. And here is an explanatory video.

Image Description: Schematic for chemical synthesis of DNA. Figure by author.

Chemical synthesis is a highly accurate method owing to this protection-deprotection cycle. Despite this, it has its limitations. It employs harsh chemicals for the deprotection steps, which can be harmful to both the handler and the environment in the long term. Purifying the sequences becomes cumbersome and costly as the sequences get longer. That’s why enzymatic synthesis becomes a compelling alternative.

Enzymatic Synthesis

Enzymatic synthesis uses a template for the nucleotides to follow. Nucleotides bases have complementary pairs: A to T, and G to C. The template will be complementary to our desired sequence. A sequence of bases called a primer will initiate this sequence formation. We also add free nucleotides with a triphosphate group attached to one end (they are called deoxyribonucleotide triphosphates, or dNTPs for short). Enzymes called ‘DNA polymerases’ will be used to facilitate the addition of the dNTPs at every step.

Shared below is a schematic for the process. If you want to dive deeper, here is a series of videos on this process.

Image Description: Schematic for enzymatic synthesis of DNA. Figure by author.

Although the chemicals used in enzymatic synthesis are less harmful, it is more error-prone than the chemical method, as we provide all the nucleotides at once, and allow the additions to happen altogether within the reaction mix.

Best of Both Worlds

A common drawback of both chemical and enzymatic synthesis methods is that we cannot easily modify or tag the DNA sequences. Tagging or modifying bases is a useful tool for storing information in DNA. But doing so might affect its coupling efficiency to the adjacent base, rendering it useless. This begs the question: is there a better method of DNA synthesis that can leverage the benefits of existing methods while mitigating their limitations?

A study published by a team of French and Swiss researchers in 2023 brings a new perspective to DNA synthesis. They build trinucleotide blocks and then arranged them one after another to build the whole sequence, instead of building the sequence one nucleotide at a time. But what is the benefit of doing it this way?

They get the best of both worlds: the fidelity of chemical synthesis and the ease of enzymatic synthesis. They used a trinucleotide, dT₃TP (deoxy tri-thymine triphosphate), as the model substrate (as opposed to the single free nucleotide used in the traditional methods). These trimers were created using the chemical synthesis method (smaller sequences ensure lower use of harsh chemicals and easier purification) and extension using these dT3TPs were done using enzymatic methods.

Image Description: Research workflow proposed by Sabat et al. 2023

They used two templates to study the activity of these trimers. Template 1 was a test template with a 15 nucleotide (abbreviated as nt) primer region followed by a series of A’s called a polyA tail (complementary to T, which is what our model is made of). This polyA tail had the space for 4 codons (total of 27 nt). Template 2 had the same 15 nt primer followed by a longer tail sequence at the end. This tail had stretches of 3 A’s (represented as dA3) sprinkled in the middle. This was a total of 47 nt. Once a successful extension was seen in the test template, they would confirm it with the second template.

When they tested their substrate with the test DNA template, the trimers were seen to not extend the primer at all, and even showed degradation. This made the researchers wonder whether the phosphodiester bonds, those which connect the nucleotides together to form the backbone of a DNA sequence, had something to do with stability. They then came up with a new model, which had phosphorothioate bonds ( the P=O structure in the phosphodiester bond is replaced with P=S for better stability against degradation). This modified trimer extended the primer completely for the test template as well as the dA3 interspaced template, proving that using codons as building blocks for oligonucleotide synthesis is possible.

What does this mean for the future of DNA data storage?

This method improves the fidelity of DNA synthesis while reducing its harmful effects on the environment as well as its handlers. It also provides an extra benefit by allowing us to modify and tag the internal bases easily without affecting coupling efficiency of the trimer: the flanking bases (i.e. bases at the end of a DNA sequence) remain unaffected.

This method reduces the time taken as well as the money spent on synthesizing DNA, which can be beneficial while designing the sequence to store data in. Reduced effort and cost can make DNA data storage more accessible and affordable to everyone.

******

Meet the Author:

I’m Saraswathi - a bench scientist by day and a writer by night. I’m currently pursuing my masters in biotechnology at IIT Guwahati, and previously did an undergrad at SASTRA University in Thanjavur. I am an aspiring science communicator. You can find my blog ‘Science Simplified’ on Medium (yes, Anagha tried to convince me to shift to Substack, but that’s for another time).

Apart from that, I love singing and listening to music (Taylor Swift tops my Spotify every single year). I also enjoy crocheting and watching sitcoms. Catch me on a random day and I’m probably scrolling through Instagram or watching The Office. I’m also a bit of a podcast nut, so if you ever need recommendations, I’m your person.

You can reach out to me at saraswathi.nvr02@gmail.com

******
If you would like to volunteer to write for the BioCompute blog (which has 577 subscribers as of today) reach out here with a brief idea of what you would like to write about, and some writes samples from the past.

I can’t guarantee timelines for now (because we are doubling down on scaling up our data storage, and we don’t have a full time community person yet), but rest assured I will get back. Any messages received via Linkedin/email for this will not be considered.

Subscribe now

At the outset it seems cool. But I dug into the research paper.

They haven’t stored 36 petabytes of data. It says they have created a medium that can store 36 petabytes, the highlight here is CAN.

They have actually stored only 156.71 KB and extrapolate the potential of their medium to 36 petabytes if they increase the length of the tape.

How does the system work?

They have created a system similar to a cassette tape using nylon.

They then create compartments in the cassette by alternating between hydrophobic and hydrophilic sections. Hydrophilic (literally meaning water-loving) regions are where DNA can be deposited. Hydrophobic (literally water-hating) regions separate the hydrophilic regions from each other, so that data from one region can be selectively read out without disturbing the others. A barcode (exactly like the ones you see on products you buy from the supermarket) is printed on each section. This acts as an address marker. Note, this is a physical barcode and not a DNA barcode, so it can be read out using an optical scanner similar to the ones used in supermarkets.

i) Writing data in

First, a short DNA sequence of 20 nucleotides (i.e. fundamental units of DNA - A, T, G, C) is immobilized in each compartment. This is referred to as a handle sequence. The handle sequence in each compartment is different.

Now the files to be stored is synthesized (using chemical methods), with an extra short sequence at the 5’ end that is complementary to the handle sequence. Complementarity in DNA is the basic principle on which a double stranded DNA is formed. An A on the first strand always binds to a T on the second (and vice versa). Similarly, a C on the first strand always bind to a G on the second (and vice versa).

So when the synthesized DNA is added to the specific section of the cassette where it should bind, the short sequence at the end (called an adapter) binds to the handle sequence. And then a PCR mix is added to extend the handle sequence so that the entire thing becomes a double stranded DNA, with the lower strand (3’-5’) bound to the nylon and the upper strand (5’-3’) bound by hydrogen bonds to the lower strand because of complementarity. PCR is a process by which multiple copies of a DNA are made. Here is an explainer video if you want to learn more about PCR.

Data in this case is stored in DNA sequences 100 nucleotides long. So if you need to store a file of 1 KB (and each hydrophilic section stores one file), this is what you would need to do. Assuming each base stores 2 bits, one 100 nucleotide sequence stores 200 bits. 1 KB = 1024 bytes = 8192 bits. So we would need 8192/200 = 40.96, i.e. 41 sequences of 100 nucleotide length to store this data. So you synthesize each of these 41 sequences with the adapter, add them to the mix, and when the cassette is dipped in the solution, they bind to the specific hydrophilic region. This is how data is written in.

The DNA is then protected by depositing a metal organic framework (of recent Nobel Prize fame) to prevent exposure to moisture and oxygen.

I am skipping the specifics of the chemicals and reactions involved here. If you are curious, feel free to check out the paper.

ii) Reading data out

In the paper, the authors have stored 5 different files in 5 compartments of the cassette. When they want to retrieve file 1 for instance, they insert the cassette into the cassette reader they have built, connect this cassette reader to their computer, and enter the file they want. The cassette starts rotating, and stops at the point where the reader identifies the right barcode for the file.

A mini-enclosure is then created to enclose that specific hydrophilic region associated with the barcode of the desired file. This is where their engineering prowess comes in. They have designed the compartment in such a way that all the reagents needed to dissolve the protective cover, separate the upper strand (5’-3’) from the lower strand are all accessible through different valves in the same fluidic system. The strands are separated by breaking hydrogen bonds between them using sodium hydroxide (NaOH). You can imagine this like having different taps connected to the same water supply in your house.

Once the lower strand goes into the solution, the cassette is removed from it, the solution separated out, DNA separated and purified from the solution, and then sequenced using a separate sequencer to . When you initially read the paper, you might wrongly assume that the reading of data is happening within the cassette-reader system. But that’s not the case. You still need a separate reader.

Which sequencer do they use?
MGI DNBSEQ-T7. This sequencer is developed and made in China. More on how this sequencer works, and how it compares to other sequencing systems in a different blog soon.

Once the upper strand goes into solution, the PCR mix is added to regenerate the upper strand by using the handle sequence on the lower strand as the template. That way, the data is regenerated each time you retrieve it. And then you put in the protective coating as before.

The authors were able to achieve 10 read cycles with 100% accuracy. They were able to store 1 text file of 0.11 KB and four image files of 51.6, 24.8, 46.9 and 33.3 KB respectively.

iii) Erasing and Rewriting data

Now what happens if you want to erase the data in a compartment and rewrite it with a different piece of data?

So here you would need to keep the handle sequence as it is, and cut out the remaining part of the sequence that has the data that was initially stored. You can use an enzyme called a restriction endonuclease to achieve this (note: all files are synthesized with a recognition site for the Mbo I restriction endonuclease, ignore this if you don’t understand what this means). Once the cutting and clearing are done, the new data with an adapter complementary to the handle sequence can be added.

The authors performed this operation once, and they had reasonable success with it.

iv) Data Reconstruction

This is done using an algorithm called DNA Fountain. I will write in detail about this in another blog soon.

A few notes on this paper:

1. They have used a 15 meter long tape to store 156.71 KB, and calculate that a 1km long tape would be able to store 36 petabytes.

2. Another claim they make is of the time taken to encode data and retrieve it - 25 minutes if you are writing for the first time and 50 minutes if you are erasing and re-writing. This is misleading, because it does not include the time taken for DNA synthesis (which was done through an external vendor), DNA purification, sequencing and computational analysis. The entire cycle would take 20 hours at the least.

3. The only real innovation in this paper is the building of the cassette and cassette reader, which offers a good electro-mechanical system to separate different files and retrieve selectively, as opposed to trying to retrieve data from an oligo pool (fancy word for a mix of short DNA sequences in a single solution). Constructing fluidic systems of this precision controlled by computers is by itself an admirable feat, but this is not an end-to-end DNA data storage system.

4. The cost of synthesis still remains a big bottleneck. I believe this is why they didn’t extend their work to the 1km tape and store 36 petabytes. If you need to rewrite data you need to re-synthesize in this case.

*********

Some updates:

1. BioCompute has won the CXXO Deep Tech Disruptor Award which comes with a 5 lakh ($5000) grant. I will be in Delhi on 30th October at the Kalaari Capital CXXO Summit to share what we are building and receive the award. I am in Delhi till 1st November EOD, if you want to catch up please reach out.
   Note: Just because you reached out, I don’t have an obligation to respond or meet. You are extending an invitation and I can choose to decline.

2. We have been offered a fully-funded opportunity to present our work at the ‘MoleculArXiv Autumn School on DNA Data Storage’ being held at the Institut d’Études Scientifiques de Cargèse (IESC) in France. We are the only non-academic lab to be selected. We are still planning our trip, if you are in France and would love to catch up, drop me a note. Especially if you are a researcher, either in academia or otherwise.

3. We FINALLY got our automation system. We are setting it up and getting it ready to zoom us into scale up.

   

   Image Description: Our intern Franci setting up the system
   

Subscribe now

It’s easy to ask why and try to pin point a single reason. But these are decisions that involve a lot of nuance and multiple factors compounding over time. While Catalog was a pioneer in developing the DNA writer Shannon, we believe that it fundamentally boils down to the unit economics of writing data into DNA not making enough sense for the market to embrace it.

That’s why we at BioCompute are focused on an approach that doubles down on bringing down the cost of writing data into DNA, bypassing the holy grail of synthesis and adopting a modification-based approach. This is what gives us the courage and conviction to keep building day after day.

**********
Some cool announcements:

1. BioCompute is now officially a member of the DNA Data Storage Alliance alongside Western Digital, Microsoft, Biomemory and more (yay!). This will help us be in rooms where computational frameworks around DNA data storage and policy recommendations are being formulated, and find relevant collaborations.

2. We are co-authoring a book titled ‘DNA Data Storage – Current Approaches and Emerging Trends’ being published by Springer Nature, alongside leading researchers in the field (P.S. this is the first comprehensive book on DNA Data Storage to ever be published). We are expecting the book to come out later this year, so stay tuned!

3. Our electronics lab is now called the Pore-formance Lab and our bio lab is called the EpiC Lounge (brownie points for figuring out why they are named so xD).

4. We are hiring a full-time ML engineer to accelerate DNA reading at our Pore-formance lab. Here is a comprehensive JD. If you love isolating signal from noise, and building ML models from scratch while working at the cutting edge of tech and with one of the smartest and most creative electronics engineers you will meet, this is the role for you. If you know someone who is a great fit for the role, do pass it on to them (there’s a surprise gift for you if we end up hiring the person you recommend).

*******
I spent most part of September in San Francisco. It’s an interesting juxtaposition of a city, but definitely one that pushes you to build (more on this in my personal blog).

Here’s the major stuff I was up to.

1. Our investor 1517 organized an Unplugged gathering of their founders, LPs and community members in the YMCA camp. It was 2 full days of being present with each other in meditation, camp fires, movie screening under the stars, and honest conversations about everything ranging from UFOs to burnouts (having 0 network helps with this). I learnt so much about company building, avoiding burnout as a founder, fundraising, and tech milestones without the shallowness of a networking event. It’s energizing to be amidst people who are building crazy stuff, while also holding on to a deep sense of compassion and hope for the world.

2. SNIA Storage Developer Conference
   This event was the anti-thesis of 1517 unplugged. Super formal, structured and lots of presentations. I got an insider view of the storage ecosystem, the wins, the problems that companies are pouring time and money into. The conference showed me that there is all the more reason to accelerate our work in the DNA data storage space given the intense storage requirements emerging as a by product of the growth in AI.

3. Crashed into a YC demo day, listened to many of the pitches and got a sense of the ecosystem

4. Visited and met interesting people at Berkeley and Stanford. I also attended the Nucleate event at Baker Labs where I met folks working on cool scientific problems in the biotech space.

5. Finally met the Savant founder Jeson Lee and his community who are chasing good quests. BioCompute was a part of the Savant accelerator earlier this year but I couldn’t make it to the demo day due to visa issues (also if you want your US Visa to get done, the best person I have found is Moiz Saifee from Saifee Tours and Travels, he sorts everything out from visa appointments to interviews and flight ticket bookings with zero hassle).

6. Visited the Founders Inc space and worked out of there for a few days. I loved the energy and camaraderie, and hope to be back there soon. I also had a candid chat with Abhinav, building Orangewood robotics, which was super insightful.

7. Identified the best automation system for us to scale up our data storage system. It’s being shipped and we hope to have it in hand by next week. A huge shoutout to Mike Ferguson for going out of his way to help me with this.

8. Attended the flagship event of Notion, with a VIP pass (thanks to Akshay Kothari being one of our early backers through the Kothari Fellowship). I loved how the event was structured so systematically and yet so informally, and how Ivan Zhao brought in Alan Key for a fun chat about philosophy and design.

   

9. Launched a community for Indian origin scientists and engineers in the US looking to move to India, or work across the US-India corridor. We are doing our first virtual meetup this weekend. If you or someone you would like to be a part of it, here’s the link to sign up.

   
   More stuff soon!

   Subscribe now

1. I was in Hyderabad from 8-10th August.
   - Spent time visiting and speaking to folks at T-Hub and CCMB. Fascinated by how the startup ecosystem in Hyderabad is evolving and how interdisciplinary they are becoming; also spent a decent amount of time chatting with scientists and engineers across domains
   
   - Visited the I-Hub at IIIT Hyderabad, and spoke to the person leading it, who has incidentally spent a lot of time at data storage hardware companies before.
   
   - Gave a little lecture to the students and faculty at IIIT Hyderabad on DNA data storage. Got a LOT of questions about how exactly this works, that got me thinking. A shoutout to Dr. Prasad and his student Ihsan at IIIT for making this happen and for giving our work a lot of visibility.

   

   Image Description: With Dr. Prasad and Ihsan, after I made them walk a kilometer to get back the water bottle I had forgotten in the lecture hall
   

2. I was in Mumbai from 14-16 August.
   - Spent some time at the Center for Development of Advanced Computing explaining our work, and discussing potential collaborations (stuff is brewing, will keep you posted)
   - Met a bunch of interesting folks from across domains interested in supporting our work with capital and networks, it’s always interesting to catch up in person instead of over a screen, some stuff under works
   - It was raining like crazy, so I largely curled up in my super comfy Airbnb reading until I had to fly back to Bangalore
   

3. My US Visa has FINALLY come through.
   I will be in SF in September (and hopefully Boston, and New York as well), let me know who are folks I should meet, what are events I should be at.
   

4. We are working on figuring out tons of stuff around lab automation, both mechanical and fluidic.
   If you work on anything closely related, do reach out for a chat. Also, do send relevant folks my way if you know excellent engineers in this space.
   

5. We are looking to chat with folks in electrochemistry.
   If you have strong fundamentals and have worked on exciting projects (regardless of industry, whether it was a personal/an academic project), do reach out.
   

6. A lot of exciting stuff happening in the DNA data storage space
   a. The Scale Lab at Imperial College London just put together an end to end integrated DNA encoding-decoding system that can be controlled using an app
   
   b. Catalogue DNA just published an interesting paper on scaling compute in DNA (which I will talk more about in a subsequent blog)
   

7. Here are some reels that made our team laugh this week, I will believe they will make you laugh too, check them out
   

Subscribe now

At BioCompute, we now have a lab of our own.

A fully-equipped (almost) 2000 square foot molecular biology lab in the heart of Bangalore.

The first DNA data storage lab in Asia, outside of an academic setting.

Image Description: Day 1 at the new office + lab. It’s interesting that the office was passed on to us from a publishing house, so much so that the space has an old world literary charm to it.

I was literally jumping around in excitement the first time Akanksha put us on Google Maps, and when I booked a Namma Yatri (my go-to ride sharing app in Bangalore) auto to our office without having to find other landmarks.

We are now all set to move ahead in full steam, scale up our tech and get to early adopters. Moving out of incubators (and home labs) gives us the flexibility to design our experiments better; build and test our hardware at a significantly faster pace.

Video Description: A sneak peak into our bio lab

********
Apart from setting up the lab, and reaching out to a lot of likely early adopters, the last few weeks have been eventful (quite literally)

1. I was at SusCrunch 2025, India’s largest and most intimate gathering for folks who are passionate about building in the climate space. I shared how the Big Pi grant that we won last year has helped accelerate our work, reconnected with a lot of entrepreneurs I look up to, made some new friends (and also won a free trip to Dancing Frog Habitat in a lucky draw hosted by Mycelium during the event).

   

2. I spoke at TEDx Bangalore about our novel approach to encoding data in DNA
   Super grateful to Sartaj, Raunak and Daniya for the opportunity, and for prepping me up for a multi-sensory talk (as Akash Kulgod describes it). I can’t wait to share the recording with you all when it is released.

   

   Image Description: If the stage looks similar to the photo above, it’s because both the events were at the Bangalore Creative Circus
   

3. My team was at the Bio Builder Meetup last weekend, where they met and interacted with a lot of folks in the biotech space and learnt a lot.

   

4. All In Capital featured us on their Elevator Pitch Season 2. A lot of folks have reached out after watching that, our team is in the process of getting back to each of you, thanks for being super patient.
   
   *******

After my last Substack post, a lot of interesting people have reached out to us expressing interest to work with us on niche problem statements. We have been figuring out how to make that work and we are onboarding our first part-time engineer this week to work on some computational aspects of BioCompute. Let’s see how that goes.

We are so blown away by the talent and enthusiasm, do keep that coming. If you are a scientist/engineer/generalist looking to build something world changing, reach out to us with proof of work. You are the best fit for us if you thrive in ambiguity, and are open to doing a lot of hard work.

******
I will be in Hyderabad on 8th and 9th August. I am already in touch with IIIT Hyderabad, T-Hub, and CCMB folks. If there is anyone else that you think I should meet do send their contacts my way.

I will be in Mumbai from 14-16th August. Help me decide where I should be and who I should meet.

Subscribe now

“It was a time when the unthinkable became the thinkable and the impossible really happened”
- Arundhati Roy, God of Small Things

This time last year, I put forth a hypothesis - that we could use DNA modification to encode data in DNA, at a significantly lower cost as compared to de novo synthesis.

It’s super exciting that we have cracked the proof of concept -we have stored data in DNA by editing in our lab, read it back using the Oxford Nanopore sequencer and fabricated our own solid state nanopores. The outcome is clear - our hypothesis stands.

This is how we currently store data in DNA

Retrieving the data we stored

The ohmic curve after we have fabricated the solid state pore

On some days when I zoom out, it feels like magic that a team of four young people got this far at the bleeding edge of tech with limited capital, infrastructure and time.

Some Wins

1. We recently wrapped up the Nucleate Activator. And BioCompute won the best pitch award in the Tech Bio category during the Demo Day. I hope to be able to go in for the Summit in August (if the visa comes through of course!)

2. I have been featured in the ‘Code of Her Own’ coffee table book put together by HerKey (India’s largest network for women professionals) in collaboration with the Karnataka Digital Economy Mission (KDEM) to spotlight women entrepreneurs building in the state. I joined the HerRising the summit yesterday in Bangalore, spoke in a panel alongside a few incredible industry leaders, and met a lot of interesting folks.

With Khushboo and Pragati who put the book together

3. Karan from Swarajya magazine featured us alongside a few incredible young founders in deep tech.

Team BioCompute celebrating the small and not so small wins

In Flux (pun intended)

We are now all set to accelerate our progress and scale up. Super excited that the 1517 fund is backing us through their Flux Capacitor program and that gradCapital is doubling down on us.

After having spent a year wobbling around between an incubator, a home lab, and a university lab (a huge shout out to Atria University for supporting us over the last few months), we are now all set to move into our own space. That means that my team is largely in flux for the next few weeks.

I am super disappointed that I have had to miss the ‘Storage and Computing with DNA’ conference happening in Paris over these few days because my visa to France didn’t come through in time. But I hope to reach out and chat to a lot of folks within the community, and build networks outside the conference as well.

*****

Over the next month, I hope to share a few open problems that we are looking for extraordinary folks to come and help us solve - either full time or part time, based on what is the best fit for both of us. If you have thoughts on what you would like to work with us on, drop me a line.

Subscribe now

Note: All of this is from the context of an early stage startup. Companies at different stages require different kinds of team members, and different work styles, and this is by no means meant to be all-encompassing. Also, I understand that it’s not easy to work at an early stage company, and just because one is not a good fit for the early stage doesn’t mean they are unemployable.

The last month was tough.

We interviewed 100+ candidates and ended up making messy hiring decisions.

Here are a few learnings from this experience.

1. Proactiveness does not always lead to successful outcomes
   
   The folks we hired were buzzing around in constant action - asking questions, coming up with ideas, tapping into their network for solutions.
   
   But the outcomes were the ultimate determinants. They could not get the most basic things done without running into some or the other issue, largely caused by a lack of understanding of fundamental principles.
   
   So our big learning is not to hinge too much on proactiveness unless it is coupled with strong fundamentals.

2. Inability to take accountability is perhaps the biggest red flag
   
   If you mess up, you own up to it, figure out why you messed up and course correct. Instead of owning up, if folks on the team come up with excuses (or worse, empty apologies mixed with justifications) for why they messed up, they already become un-coachable.
   
   It’s nearly impossible for un-coachable people to solve hard problems because they are either unaware of what they don’t know, or are covering up what they need to learn. Being open to learning (not just saying you want to learn, but showing you can learn) is a pre-requisite to solve hard problems, and in turn thriving at an early stage startup.

3. Response to failure is often a good marker for success in a startup
   
   I understand that a lot of these responses are shaped by what kinds of environments someone grew up in. Fear and shame are often natural responses to failure if you grew up in a family or society that valued you only when you succeeded.
   
   However, an individual’s ability to succeed in uncertain environments is determined by how conscious they are about these patterns, and how they are working on reframing their failures within their own minds. I know this is hard hard work. But if people are not willing to/able to work on themselves, then perhaps it’s hard to expect them to commit meaningfully to a long term vision.
   
   I am cognizant of the fact that a lot of these responses are skewed based on the condition of the job market, the kind of privilege you have or don’t. But we are all trying to succeed in a capitalist world (which unfortunately rewards and amplifies privilege), and as a startup we need to maximize our odds of success by hiring the best folks.
   
   Ideally don’t hire people who have a propensity to panic in the face of failure; they need more work on their nervous system before they can be a good fit for an early stage startup. We are figuring out how to determine a candidate’s response to failure during the interview process - would love to get thoughts from other founders/hiring managers who have cracked this to some extent.

4. Ability to leverage available resources is an underrated skill
   
   Testing for whether someone is resourceful enough, i.e. when they run into a problem do they just assume that nothing can be done, or do they find their way into plausible and feasible solutions, using whatever is available to them (including google/chatgpt/reddit), is an important aspect. So now we have started testing for this in our technical rounds.

5. Don’t hire people who just want a job.
   
   This is so hard to say out loud because there is such systemic inequity within the candidate pool. However, an early stage startup cannot afford to hire people who don’t have an innate hunger to solve a problem (I am not referring to a societal problem here, but a problem in the sense that it is intellectually challenging to get to a solution).
   
   Just because someone has the skillsets or experience required to solve a problem, they need not be able to solve it, because they don’t spend enough time worrying about it, and thinking about it from the right first principles framework. It’s hard to determine why a candidate applies to your company. The obvious answer is that they are looking for a job that pays. But we believe there are ancillary markers for what matters to them (like why they picked the projects they picked in the past). If they are only looking for something that gives them a monthly salary and nothing else (which is not wrong on any level at the candidate’s end; it’s probably the most peaceful way to operate in the world), they are perhaps a terrible fit for an early stage company.

6. Hire people who invest time in activities outside of work
   
   This perhaps sounds counter-intuitive. But we have grown to realize that to bring your fullest creative self to work, you need to have healthy avenues for delight. You don’t need to be an expert at things that give you joy, but the fact that you invest time in diverse things enables your neurons to fire differently, and these can in the long-run translate into connecting dots that others haven’t connected before. Folks who can do that are a valuable asset.

1. This list is definitely going to be work in progress (as a mentor in the biotech space who has been hiring for the last two decades told us, his success rate is still 70%) but I think it is definitely worth putting it out here to track the evolution of our hiring process.
   
   I would love to connect and chat with founders building in deep tech and figure out how you are setting up hiring pipelines and what metrics you use for decision making.

   
   

   Subscribe now

Image Description: Our anniversary celebration
P.S. A shout out to Olive Drops for hosting us, and Chiris Patisserie for the cake (check both of them out whenever you are in north Bangalore next)

The last year has been a rollercoaster for me - getting the first cheque for BioCompute in my final semester of college, graduating, moving to a new city, setting the company up from scratch, hiring folks, putting processes in place, procuring raw materials, sorting out (a LOT of) documentation, doing tons of virtual meetings, creating a community of young biotech founders, travelling for work, juggling time with friends and family (most of whom have been extraordinarily patient) and more.

Everyday is a new challenge, and it is not very often that I sit down and take stock. But this was one of those days that I did. I realized that the company has come so far - from being an idea on a google doc to a 4-member team working on it full time, equipment and (an albeit makeshift) office and lab to work with/in. Super grateful to everyone who has backed us with capital, networks, mentorship and (most importantly) emotional support on days that looked the bleakest.

Website Launch

In the honour of our first anniversary, we launched our brand new website that is now accessible at www.biocomputeinc.com.

Image Description: Snapshot of the website

Shifting the domain name from substack to our new host platform, we ran into a few hiccups, which took us nearly 15 days to resolve (hence the delay in putting it out here). Super grateful to Rakshith for spending some time over a weekend helping us troubleshoot.

This is V1 of the website, we are open to any and all feedback to make it better.

Meet Naveen

The man behind our new website (it’s also interesting that his name translates to meaning new/fresh in Sanskrit) and our in-house electronics engineer building solid state nanopores to read DNA.

As a first generation college student from Coimbatore, Naveen is majoring in electrical and electronics engineering (yup, he is still in his final semester). He has previously built a smart motor management system to help businesses save on maintenance cost, kick started the electric vehicle club in his college to compete in national and international level competitions and built a project management system from scratch for his EV club.

In the last two months that he has been at BioCompute, he spent a lot of time learning about nanopores, designing PCBs, running some initial experiments with our existing setup, and then picked up website development as a side quest (If the side quests of our team members come out this cool, imagine what the outcome of their main quests would be).

Image Description: Naveen learns the ropes (or rather screws) of the trade from the rest of the team

Super grateful to Shrirang for helping us with technical interviews, that ultimately led us to bringing Naveen on board.

Here’s to gearing up for the next phase!

Subscribe now

“Mad Hatter: “Why is a raven like a writing-desk?”
“Have you guessed the riddle yet?” the Hatter said, turning to Alice again.
“No, I give it up,” Alice replied: “What’s the answer?”
“I haven’t the slightest idea,” said the Hatter”

- Lewis Carroll, Alice in Wonderland

Over the last few weeks, I have been noticing that I pretend a lot. I pretend to know things that I haven’t the slightest idea about - and surprisingly enough, I do this often without noticing it in the moment only to regret it later.

So, I dug into this a little deeper and I figured out that this pretense often shows up with topics that I have heard of and read about at the surface level (and sometimes I want to believe I understand because of how often certain words are thrown around), but don’t fully understand how it works. And now I am on a conscious quest to rework this pattern.

Baby Steps

As the first step, I read my first full paper1 on deep learning (with ample help from Gemini and GPT Deep Research) and how it is being leveraged to store and access data stored in DNA with minimal error rates.

The title of the paper is ‘Scalable and robust DNA based storage via coding theory and deep learning’. Sounds (unnecessarily) complicated. I started by breaking this down.

Scalable and robust DNA based storage is fairly straightforward. It refers to storing binary data in DNA in a manner that enables us to retrieve it without losses or errors (hence robust) and can be leveraged even for larger amounts of data than what has been demonstrated in the paper (hence scalable).

Now what is coding theory? It is a fancy word for a broad set of concepts that enable encoding, decoding and transmission of data in a way that minimizes errors, most often by introducing redundancy. Redundancy refers to adding some extra information to your existing data in order to enable retrieval with minimal errors.

And what is deep learning? This is so much of a buzzword that we all think we know what this means. But I still want to break this down.

Let’s start with what machine learning is. It refers to algorithms that enable computers to identify patterns from pre-existing data that it can now apply to newer data of the same kind. Deep learning is a type of machine learning wherein the process of identifying patterns (i.e. learning) happens across multiple layers of an algorithm, with each subsequent layer becoming more adept at picking out more abstract features than the earlier one. It is because there are several layers that this kind of ML is referred to as deep learning or deep neural networks (DNNs).

The Method in the Madness

The authors of this paper are attempting to retrieve data stored in DNA with reasonable accuracy while minimizing the amount of compute they would need. They call their system ‘DNAformer’. Here is a schematic of how it works.

“Though this be madness, yet there is method in't.”

― William Shakespeare, Hamlet

A few key takeaways:

1. There wasn’t enough data to train a deep neural network on error correction associated with DNA synthesis and sequencing (using Illumina and Oxford Nanopore).

In order to solve this, training data was generated based on a general understanding of synthesis and sequencing errors from a tool called SOLQC which is used for the quality control of synthetic DNA, i.e. it has all the parameters in place for error determination. The important point to note here is that the model isn’t based on identifying specific characteristics such as file structures or semantics of sequences, it is purely trained on filtering out specific kinds of errors that occur frequently, thus enabling the model to be used across structured and unstructured datasets across the board.

This could be a useful approach for generating synthetic data sets for use cases where there isn’t enough data to train models, and it’s too expensive to generate data to train models. However, I am unsure how the reliability of synthetic data sets can be determined, apart from the obvious step of using that to train a model and then testing the efficacy of the model on a known data set. If you have any resources on this, please send them my way.

2. Error correction is not an afterthought (unlike what I assumed)

Indices (or address markers) are added to the sequences to enable them to be sorted into clusters later on. Physical constraints, such as ensuring that the GC content of a DNA sequence is nearly 45-50% to ensure stability of the molecule, are accounted for before the sequence is generated.

3. Speed is prioritized.

Here is what it does as soon as sequencing data is obtained.

a. Clustering
The different DNA sequences are grouped into different clusters based on their indices. This is done fairly quickly based on a sorting algorithm approach, but is likely to be ridden with errors which the subsequent steps (aka the DNAformer) are well poised to correct.

b. Reconstruction
This is where I was stuck at for a really long time as I made my way through the paper. I will try to break it down in as easy a manner as possible.

The reconstruction step has three components:
i. Convolutions: These are used to align the sequences and prepare the data for the transformers (that come in the next step). The alignment and embedding modules in the figure above are convolutions.

ii. Transformers: These are used to find long range dependencies in the sequence (eg: how a small part of the sequence at the beginning of a long strand affects a small part in the middle or the end of the sequence)

iii. Confidence filters: Confidence values are assigned to each cluster based on the predictions of convolutions and transformers. This is expressed as a mathematical equation involving the sequence length and expected probability of a nucleotide (A, T, G or C) at a given position.

The confidence filters reject clusters that have a low confidence value and a low cluster size (i.e. number of DNA sequences that got grouped into a given cluster) but wants to get a second opinion on clusters that have a low confidence value but a large cluster size (thresholds for cluster sizes are already set).

The second opinion comes from the CPL algorithm, which is basically a probability based algorithm that compares every sequence to every other sequence and determines where errors could be. Instead of sending every sequence obtained from the sequencer to a sequence alignment algorithm (which does sequential processing as opposed to parallel processing, the latter being faster), this mechanism only sends those clusters that require a second opinion. That improves speed.

c. Decoding
This is straightforward. The sequences that come through the reconstruction step are then converted back to binary based on a converse of the encoding algorithm and error correction codes using a tensor product algorithm. This algorithm has redundancy across two dimensions, and hence enables efficient error correction.

******
That was quite a huge info dump (if you are as new to concepts in deep learning as I am), but attempting to explain concepts to others helps me learn better and I am glad you stuck around. Let me know which parts of these made sense and which ones didn’t, happy to attempt those again.

If you are an expert in the deep learning space and find that I have interpreted concepts incorrectly, do reach out and help me unlearn.

P.S. Shoutout to Ben for helping me access the paper which was originally behind a paywall. If you are curious but don’t have direct access, feel free to reach out I can pass on a copy to you.

Subscribe now

Image Desc: At Vigyan Lab’s sustainable micro data center

Here are some of our key takeaways:

1. Space and energy consumption of data centers are problems that companies care about for two reasons:
a) because it affects their bottom line
b) their clients (especially those in Europe) require it for scope-2 and scope-3 reporting

2. There is a barrage of software and software-hardware-integrated companies being built around helping other companies optimize their data center (and IT) costs, and they have a reasonable number of paying customers and order volume for their tools.

3. There is a reasonably big market in India for archival data storage.

We realized that all of the folks we have been talking to and leveraging to understand the demand for archival data storage have not been the real decision makers. We had eliminated research institutes and the likes from being potential customers, but we met companies selling to these customers and being paid for their services. We found some interesting leads to explore on these fronts.

4. An alternate movement for sustainable and locally built IT infrastructure is evolving in India, both from the technology and policy lens spearheaded by organizations like the Green Compute Foundation and micro data center providers like Vigyan Labs.

The storage and compute costs of big players like AWS and Azure are often too high for companies in India. Research institutes often cannot tap into these because funding agencies do not allow them to utilize grants from tax-payer money to pay private cloud providers. Hence, there are many opportunities to build compute architecture in India for India.

5. There are challenges to switching from established to local players in the data storage market
a. There are upfront switching costs that need to be paid if you are switching out of an Azure or an AWS
b. Even if a company wants to prioritize a local data center provider over an established player, the customers of the said company often don’t trust these local players. Hosting on AWS or Azure tag gives some degree of credibility to the company in the eyes of the customer. The reliability (or the perception of reliability) of local data center providers needs to be established for them to compete with AWS or Azure in India. I talk about perception because some of these local data centers have shorter downtimes than AWS or Azure servers do for a given amount of data.

The little (big?) surprise

That being said, we are currently signing up companies for early access to our DNA data storage.

If you are a company that currently does on-premises storage of archival data, or a data center service provider that wants to explore opportunities to replace your archival data storage systems with a more space and energy efficient one, hit us up. The first 5 companies to sign up with us gets a life-time special pricing offer.

Happy to share more details one-on-one.

Subscribe now

We FINALLY have our very own ONT nanopore sequencer to run some experiments. It has taken us nearly 8 months (and tons of emails) to get our hands on the ONT.

When I first laid my eyes on our sequencer, my first thought was gosh isn’t it beautiful? I have never felt about any other electromechanical device (including phones and PCs and headphones that a lot of folks fuss over), and this one is special not just because it reads DNA. It is perhaps the first device that has protein structures interfacing with electronic components, and is the most modular of them all.

Image Description: The ONT nanopore sequencer

Super grateful to Yashas, Anulipi, Saurabh, Kishan and Sachit for helping us with the ONT procurement.

I have a lot to say about how ONT isn’t the best nanopore sequencer out there (although it’s the only company in the world that sells this) and how we need to build our own solid state pores to ensure that flow cells don’t run out so quickly. But that’s for another day. For now we are admiring the gorgeous engineering of our pocket-sequencer (which a good friend described as ‘straight out of a James Bond movie’) and getting ready to run some experiments.

The Home Lab Era

“Freaks become norms, and norms become extinct. Monster by monster, evolution advanced”
― Siddhartha Mukherjee, The Gene: An Intimate History

I believe that this quote, while attributed to evolution in nature, also fits in well with the evolution of a company.

We have set up a home lab to run our bio and electronics experiments in tandem. It’s exciting to graduate from having just a bench in an incubator to having an entire (albeit hacky) room to try out experiments, build, break, rebuild and get to our eureka moments. It’s also funny that the first ever fridge I bought is for BioCompute, and not for my home (yeah I do have a long standing debate with friends about why I don’t ‘really’ need a fridge at home).

Video Description: A sneak peak into our home lab

We still rely on the incubator for bigger and more complex equipment but a home lab helps us iterate and get to our POC faster. We hope to be one of the first freaky bio/climate tech unicorns to spin out of a garage living room (but you get the point).

Super grateful to Zayner and the entire ODIN team for helping us source a lot of our equipment and consumables, and for being an all-round inspiration. And to Nitish for lending us a vortex at such short notice.

Image Description: Ajay bidding farewell to the first ever pipette tip used at BioCompute

PCB Design

We are graduating from breadboards to PCBs for our nanopore fabrication circuit. Akanksha is hard at work designing the PCBs, identifying the right manufacturer and getting it out as soon as possible.

Super grateful to Prithviraj Narendra from the Hardware Mafia community for helping us troubleshoot our PCB, and to Anumol Dominic at IISc for sharing a bunch of insights. If you have leads on really good PCB manufacturers in India (preferably Bangalore), do send them our way.

We are also onboarding an electronics engineering intern very soon, stay tuned to meet our new teammate!

Nucleate Activator

This week we received our acceptance to Nucleate, one of the best no-strings-attached biotech accelerator programs in the world

Image Description: The Welcome Email to Nucleate

For the next few months we will be working closely with mentors on different aspects of the company from setting tech milestones to IP strategy, team management and more. Super excited to join this cohort, and grateful to Heer for encouraging me to apply.

Market Deep-Dive

Akanksha and I will be at the launch of the Sustainable IT Center of Excellence in Mysore, where we will be speaking to a range of folks in the data center ecosystem, who look at storage and compute from a sustainability lens.

If you are in Mysore and would like to catch up/have recommendations for what we should be doing in the city (including and beyond networking) hit us up.

Subscribe now

Image Description: DNA book by Asimov Press in collaboration with a few DNA storage companies

Of course, the average consumer cannot read the DNA sequence within the comfort of their homes like they would a physical or a kindle version of the book. And that’s partly why we at BioCompute are so keen on building out the entire read-write-store-retrieve stack in a modular hardware, apart from the fact that a piece of hardware like this can significantly reduce the space and energy consumption of data centers.

Where are our experiments at?

1. Nanopore fabrication (to read DNA)

We have finally managed to get the circuit for dielectric breakdown up and running (a huge shoutout to everyone who helped us troubleshoot). We naively assumed that this means we can just insert our chip into the flow cell and make our pores.

But while attempting pore formation, we accidentally broke the screws of the cassette and spent a week trying to procure screws that match the dimensions. It didn’t help that we were attempting to do this while the rest of the world was taking time off to celebrate Christmas and New Year. Finally, Akanksha and Ajay made their way to SP road (nearly 18 km from our workspace, but dreaded because of the infamous Bangalore traffic) to get custom-made screws that fit into our flow cell.

Image Description: The notorious screws

Now that screws aren’t screwing us up anymore, we set out to break down some membranes. However, we are running into very high leakage currents which should ideally not happen in the kind of system we have set up.

Image Description: Our leakage current is 10x higher than what it should be to make pores

We are now in the process of troubleshooting this so that we can get to making pores and optimize the process to get to reasonable repeatability. We can then use these pores to read DNA. And we are hiring a full-time electronics engineer to accelerate progress, if you are excited about this do check it out and apply.

2. Enzyme synthesis (to edit DNA)

We came up with the sequence for an enzyme to edit DNA at the single nucleotide level, and have been working closely with a CRO to express and purify it so that we can determine the efficacy of the enzyme.

While the protein is being expressed fairly well in our expression systems, it unfortunately isn’t soluble enough for us to purify it (and we need to purify to be able to carry out experiments outside cells).

We are now attempting to express the protein in a very specific bacterial strain that is optimized for disulphide bonds to improve solubility. This is an important problem to solve, and also a very important piece of our IP as we take DNA storage to commercialization.

We look forward to chatting with enzyme engineers and protein biochemists to get their insights on optimizing the expression evaluation as well as optimizing the enzyme in the long run (thanks to Sujoy and Vaibhav for chatting with me in the last week).

3. I am attending the Accelerating Biology conference by the Centre for Development of Advanced Computing (CDAC) in Pune from 7-9 January (and writing this during the inaugural ceremony, because why not). I am excited to connect and chat with folks at the intersection of data and biology to gain insights and explore potential collaborations.

*******

A few things I have learnt from the last 6 months of being a full-time founder.

a. People are more generous than you think they are
The biggest learning has been the perspective altering (and hence a tad bit disappointing and a lot more humbling) realization that none of us can operate independently. This is starkly visible to you when you start building a company. At all points, you find that you need things that cannot be done in-house.

When you are trying to grapple with that fundamental truth and live with it, that’s also when you realize that people are more than willing to support. In the last few months alone, the number of people who have opened up their purses, homes, networks and calendars to help us is incredible. A lot of times these have been people who barely know us, but were fascinated by what we are building and stepped up. There are people who think of me when they come across a research paper, a blog article or a book, and then very generously send them across. I am incredibly grateful to these people and am trying to pay it forward as much as possible.

b. Building a company culture while optimizing for optimal outcomes (constrained by time and money) is hard.
It’s easy to be a rebel, but setting up systems that work (and dealing with other rebels) is a tough iterative process. I am still figuring out how this works.

c. A good Canon printer costs only INR 2,600 on Amazon
I have grown up believing that a printer is a super expensive purchase. I was pleasantly surprised when I found out what it costs today - perhaps the cost of printers themselves have come down, or our purchasing power has gone up. But anyways, I wouldn’t have checked if I hadn’t started a company.

d. Getting into a disciplined lifestyle is hard but is worth doing
Being regular to morning workouts, therapy sessions, and having a mostly fixed sleep schedule is helpful to gather and sustain energy for decision-making and then the drudgery of execution.

e. Looking for delights is a meaningful pursuit
Ross Gay’s conversation with Krista Tippet in this episode of ‘On Being’ and his book ‘The Book of Delights’ gently nudged me to actively look for delights in everyday life, and be generous with what you qualify as delights. Given how our brains are wired to pick out threats, being conscious about joy is perhaps the only way to invite it in. This practice has been extraordinarily valuable on days when it feels like BioCompute (the scope and scale and the everyday challenges) is crushing me.

f. Communities and supportive friendships are extraordinarily valuable
It takes a village to raise a company; more precisely a village of equally ambitious and irrationally optimistic folks. Finding them and hanging out with them keeps you going. I am finding and building my own Lunar Society (thanks Amruth for sharing this one, and excited for the community you are building out).

Subscribe now

1. Akanksha pitched BioCompute at the regional finals of EO GSEA in Noida and secured the first place for us, with a cash prize of 2 lakhs. Super pumped for the next set of rounds!

2. Nucleate, the largest global community of bio-innovators, just launched their India chapter.

   
   I had the opportunity to attend their inaugural community meet up in Bangalore. It was incredible to see folks from academia, startups and the investing ecosystems get together and discuss collaborations. A huge shoutout to Heer (also a BITSian) for introducing me to Nucleate and keeping me posted about all the cool stuff they do. Excited to see how the movement pans out in India.

3. I was at the Emergent Ventures ‘un’conference in Kochi from 9-11 December. It was a life changing experience to be in a room with nearly 150 people who are shakers and doers, building some of the best stuff within their respective domains. EV is a truly no-strings-attached grant that is given out to folks who want to work on some exciting projects.
   
   The best part of the conference was when three of us from different age groups and backgrounds - Shrirang (computer science and math), Vedanth (biology) and I (chemistry) sat down for nearly 2 hours working out how a bio-based system could be used to solve nth order polynomial equations. I am looking forward to initiating collaborations with folks I met at the conference to drive BioCompute forward.

   

4. Our R&D is slowly taking shape and coming along. It’s a lot of experiment, troubleshoot and repeat. We will keep you updated when we get to some crazy insights.

A few interesting updates from the DNA storage space:

1. French startup Biomemory building DNA data storage systems raised $18 million in series A last week. It’s exciting to see that the molecular storage space garnering so much interest from the market, and solidifies our conviction in getting DNA storage to commercialization fairly soon.

2. Two intriguing papers came out earlier this month.
   a) High speed sequential DNA computing using a solid state DNA origami register1

   

   Using strand displacement reactions to do DNA computing has been around for a while, but what’s uniquely important about this paper is that the scaffolding has been done using DNA origami (a composite structure made up of DNA strands in different orientations, pretty much like paper-fold origami) which is on a solid state nanopore. This brings together the processes of reading and writing data in DNA as opposed to operating them as separate systems. Thanks Vaibhav for sending this paper across.
   

   b) Post translational digital data encoding into the genomes of mammalian cell populations 2
   This is not directly related to what we are building at BioCompute but is exciting nonetheless. Here, the authors use a self-targeting CRISPR Cas9 enzyme to create double stranded breaks in mammalian cells and then looks at how different cellular conditions or stressors incorporate different nucleotide sequences; enabling the cell to act as a storage system for these cellular conditions. The interesting bit is that data gets encoded in the overall distribution of bases as opposed to defined sequences of nucleotides.

   
   I believe that this would be an interesting domain to explore when it comes to bringing down the cost of writing data into DNA. Thanks Yashas for sharing this one with me.

   
   

   Subscribe now