Healthcare IT · Scientific data infrastructure
TetraScience
Software that pries laboratory data out of instrument vendor silos and reshapes it for machine learning. Founded in 2014 as a lab hardware company, rebuilt in 2019 as a cloud platform, and now selling an AI program to large pharmaceutical R&D organizations.
A glass thermometer and a shout
The experiment was electroplating: coating one metal with another. Alok Tayi, then a postdoc at Harvard, called it “quite literally a centuries old experiment.” Running it took two scientists two hours.
“I had to sit there with a glass thermometer, monitoring the temperature of the chemical experiment and I had to scream it out to my colleague who wrote down the data points in pen and paper,” Tayi told an interviewer in 2017.
That was February 2014, and the comparison that came to him was domestic. “We said when we have capabilities like Nest, the smart thermostats in the home, the function and the value of that same capability in the laboratory however is literally 100 times higher.”
He had dinner that evening with Sal Savo, a Harvard optics researcher who had built scientific hardware. Later that quarter he met Spin Wang, an MIT graduate student who had already wired together more than ten instruments in a Cornell radio frequency lab because recording the numbers by hand wore on him. The three incorporated TetraScience at the end of 2014 and went through Y Combinator in the summer of 2015.
The data does not move
A large pharmaceutical research site runs hundreds or thousands of instruments: plate readers, mass spectrometers, chromatography systems, flow cytometers, freezers. Each writes in its own format, into its own software, often onto a machine in the room. Moving a result into an analysis is frequently a person retyping numbers.
“We felt the pain of manually tracking data and not having a consistent interface for all our equipment,” Wang told MIT News in 2018.
The industry names the bill for that friction after itself: Eroom’s Law, Moore’s Law spelled backwards, the observation that the cost of developing a drug has roughly doubled every nine years. TetraScience did not invent that framing, and it is not alone in selling laboratory data software. Its claim is about order of operations: data has to be freed from vendor formats and rebuilt into shared schemas before a model can use it.
The company that was rebuilt
The first TetraScience was a hardware business, shipping an internet-of-things hub, a modified Raspberry Pi at the start, that plugged into freezers, incubators, scales and autoclaves and pushed readings to a cloud dashboard. An early trial at the MIT Media Lab monitored beehives. By January 2018 the company told MIT News that more than 70 major pharmaceutical and biotech firms used the platform, and that it had reached more than 60 percent of the top 20 pharmaceutical companies.
Then it changed shape. In an April 2021 post, Patrick Grady dated the current business to May 2019, when he and Wang, he wrote, “completely reimagined and rebuilt the company”. The hardware line was not carried along. In December 2021, Elemental Machines, a Cambridge company the release described as a former competitor, bought the Tetra Lab Monitoring business outright: hardware, talent and instrument integrations. Savo went with it and became Elemental Machines’ chief technology officer.
Money, then AI
Eight months before that sale, in April 2021, TetraScience announced an $80 million Series B co-led by Insight Partners and Alkeon Capital. The month before, Novo Nordisk had put out a joint release naming TetraScience as the platform for automated instrument data acquisition in early research. In June 2021 the product went into the AWS Marketplace.
The product story then turned toward AI. A collaboration with NVIDIA was announced in November 2024, built around BioNeMo models and a cell line selection application. In October 2025 the company launched a program it calls the Scientific AI Lighthouse, or SAIL, with Takeda as founding partner. In January 2026, timed to J.P. Morgan healthcare week, it announced a collaboration with Thermo Fisher Scientific, and in 2026 it opened a European headquarters on the Novartis campus in Basel, with Wang relocating there.
On February 13, 2026, the company sold $14,999,985 of equity to eleven investors. It filed the notice two weeks later. It did not announce the round.
What is proven, and what is still claimed
| Evidence | What the record shows | Source type |
|---|---|---|
| Regulatory and clinical | No FDA 510(k) or PMA records, no registered trials, no NIH or NSF awards found as of Sep 22, 2026. | Public record |
| Named customers | Novo Nordisk, joint release Mar 23, 2021. Takeda, founding SAIL partner Oct 23, 2025, with two executives quoted by name. Alexion, cited for outcomes on the company site. Merck & Co. and SmartLabs, named by VentureBeat as 2020 additions. | Company and partner |
| Customer count | More than 70 pharmaceutical and biotech firms, and more than 60 percent of the top 20 pharmaceutical companies, both attributed to the startup by MIT News in January 2018, before the lab monitoring line was sold. No comparable current figure found. | Company-stated, 2018 |
| Revenue | No dollar figure disclosed. The company reported 10x revenue growth in 2020; VentureBeat called it a 1,000% uptick in annual recurring revenue. Both Form D filings mark the revenue range “Decline to Disclose”. | Not disclosed |
| Vendor partnerships | AWS Marketplace listing Jun 2021, NVIDIA Nov 2024, Thermo Fisher Scientific Jan 2026 as covered by Fierce Biotech. Databricks, Snowflake and Microsoft named by the company. | Independent and company |
| Performance numbers | A March 2026 Databricks post reports 80% less review time at “one global biopharma”, 94% binding prediction accuracy in a product “deployed in production at a top-20 pharma”, and cell line work cut from 8 months to 2.5. No customer named. | Partner blog |
| Capital on file | $50,999,950 sold in the April 2021 notice and $14,999,985 in February 2026. The Series B was announced as $80 million the same day the 2021 notice was filed. | Filing differs from release |
| Who founded it | MIT News in January 2018 names three co-founders: Wang, Savo and Tayi. Grady’s own April 2021 post dates his work with Wang to a May 2019 rebuild. He is currently listed as founder, chairman and CEO. | Accounts differ |
| Headcount | Revelio Labs estimates 229 employees in Q2 2023, a low of 162 in Q2 2025, and 235 in Q1 2026. No layoff announcement found. | Third-party estimate |
Read plainly: the customer evidence is real but thin, and the strongest numbers about what the software does come from a partner blog that names no customer. Takeda is the one large pharmaceutical name attached to the AI program with quoted executives behind it. The most specific customer counts on record date from 2018 and describe a product line the company no longer owns.
What to watch
- Whether an announced round follows the quiet $15.0 million equity sale filed in February 2026.
- A second named SAIL partner. Takeda is the only one so far.
- A current customer count, to replace figures published in 2018 about a divested business.
- Named customers, or published work, behind the March 2026 performance claims.
- Whether headcount holds above the 2025 low as the Basel office staffs up.
In their words
“I had to sit there with a glass thermometer, monitoring the temperature of the chemical experiment and I had to scream it out to my colleague who wrote down the data points in pen and paper.”
Alok Tayi, co-founder, The Hardware Entrepreneur, 2017 · Interview
“our vision is to connect every single scientific experiment and instrument to a single online dashboard”
Alok Tayi, co-founder, The Hardware Entrepreneur, 2017 · Interview
“Our technology is establishing a ‘data highway’ system between different entities, software and hardware, within life sciences labs.”
Spin Wang, co-founder and CTO, MIT News, 2018 · Independent
“Inbound interest in the Tetra R&D data cloud from global pharmaceutical companies has increased dramatically, despite very little investment in marketing.”
Patrick Grady, CEO, to VentureBeat, 2021 · Independent
“We’re witnessing the rise of a $300B industry data cloud.”
Deven Parekh, managing director, Insight Partners, Series B release, 2021 · Company release
“I believe that this acquisition will result in many upsides for TetraScience Lab Monitoring customers who will benefit from both companies’ investments of the last several years.”
Salvatore Savo, co-founder, on the sale of Tetra Lab Monitoring, 2021 · Partner release
“Embedding AI and digital technologies across the R&D value chain is one of Takeda’s core strategic areas for our future.”
Nicole Glazer, Head of R&D Data, Digital and Technology, Takeda, 2025 · Company release
“At the time, our hypothesis was highly non-consensus. Some called it naive and even misguided.”
Patrick Grady, founder and CEO, company essay, 2026 · Company
Related companies
Sources
- Public recordForm D, TetraScience, Inc., CIK 0001855742
- Public recordForm D, TetraScience, Inc., CIK 0001855742
- Public recordImproving protein therapeutic development through cloud-based data integration
- IndependentStartup makes labs smarter
- InterviewConnected lab, with Alok Tayi of TetraScience
- IndependentTetraScience scores $80M series B investment
- IndependentTetraScience raises $80M to help life sciences companies analyze data
- IndependentThermo Fisher Scientific teams up with Nvidia, TetraScience
- IndependentTetraScience employee count and headcount data
- CompanyTetraScience Secures $80 Million Series B
- CompanyAnnouncing Our Series B: The What, When, Why, Who, and Where
- PartnerNovo Nordisk and TetraScience partner on a cloud-based instrument data platform
- PartnerElemental Machines Announces Acquisition of TetraScience Lab Monitoring
- CompanyTetraScience Launches Scientific AI Lighthouse Program with Takeda
- PartnerHow TetraScience accelerates biopharma with production-ready data
- CompanyReimagining Science for the Era of AI
Profile researched and written by Healthcare Discovery. Last updated September 29, 2026.
