Therapeutics · AI drug discovery
Reverie Labs
A Cambridge company that used machine learning to design brain penetrant kinase inhibitors for cancer. It raised more than $32 million, signed Roche and Genentech, never put a molecule into a person, and sold its software to Ginkgo Bioworks in February 2024 for about $2.4 million in stock.
A better coin mint
Eleven months after the sale, the co-founder who built the software explained why it had not worked, using a metaphor about coins.
Drug discovery, Ankit Gupta wrote, is an unfair coin. Flip it and about one time in ten you get a molecule with clean animal data and a clean first look in people. Venture firms grade coins well and buy enough that a few land heads. What Reverie Labs tried to build was not a coin but a mint: a machine that stamps out better coins.
Proving a mint works is the trouble. “Unfortunately, there is not any convincing way to measure how good our coin mint is other than minting a coin and flipping it.” And, he added, “with the funding that most techbio companies have, we had the funds for essentially 1 coin flip.”
Reverie never got to flip it.
The gap it was aimed at
Kinases relay growth signals inside cells, and when they misfire they drive a great deal of cancer. Dozens of kinase inhibitors are approved drugs. Most cannot cross the blood brain barrier, which is why brain tumors and brain metastases stay among the hardest cancers to treat.
That gap was the target. Reverie's pitch was that kinases suit machine learning unusually well: hundreds share structural motifs, and decades of published binding data give a model something to learn from. Train on potency, on selectivity across the kinome, and on the properties that decide whether a molecule survives a body, then generate millions of structures and rank them.
The first half worked. By Gupta's account, “for every one of the ~10 programs we worked on in the life of Reverie, we had multiple potent, selective inhibitors within the first 2 months of starting the programs, and then had to spend months to years optimizing ADMET profiles.” Binding was the easy part. Making a potent molecule behave like a drug was not, and the models were data starved on exactly the properties that mattered.
Two Harvard roommates
Jonah Kallenbach and Ankit Gupta met on their first day of college at Harvard and both finished computer science degrees in 2017. Neither had run a drug program. They started the company that year, named it after a detail in the HBO series Westworld, and got into Y Combinator's Winter 2018 batch. Accounts of the founding team differ: TechCrunch in March 2018 named a third co-founder, the MIT graduate Connor Duffy; Tracxn names a third co-founder as Hirsh Jain. Kallenbach was chief executive throughout, Gupta chief technology officer.
The original business was software, and it did not survive contact with the buyer. Pharma does not buy much software, Gupta later wrote, and the people who design molecules do not hold the software budget. Counted from the bottom up rather than the top down, “Suddenly, the TAM looks more like double digit millions instead of billions.”
Two pivots and a partner
Reverie pivoted at Demo Day to selling molecules instead of the tools that made them. In July 2020 that produced the high water mark: a multi-target collaboration with Roche and Genentech, with an upfront payment, preclinical, clinical and regulatory milestones, and tiered royalties. No figure was ever disclosed. In February 2021 Ridgeback Capital Investments led a $25 million Series A, bringing “the total capital raised by Reverie to more than $32M.”
Then came the second pivot, to owning its own drugs, and that is where the money ran out. Three years later Reverie sold the software that had made them.
What is proven, and what is still claimed
| Evidence | What the record shows | Source type |
|---|---|---|
| Registered trials | No study on ClinicalTrials.gov lists Reverie Labs as sponsor or collaborator, checked Sep 23, 2026. | Not found |
| Pipeline as published | The pipeline page archived Feb 4, 2024 lists RVL-101 as a “CNS-penetrant CDK4/6 inhibitor” in cancer, plus RVL-102, RVL-103 and undisclosed Roche and Genentech candidates. Its stage bars run Early Discovery, Late Discovery, IND-Enabling, Clinical. RVL-101 sits at 69 percent of that scale, inside the IND enabling band; the other two at 38 percent. Nothing reaches Clinical. | Company-stated |
| Independent database | Patsnap Synapse lists three drugs associated with Reverie Labs, Inc., highest phase Preclinical, one target pair recorded as CDK4 x CDK6. | Independent |
| Roche and Genentech deal | Announced Jul 8, 2020, multi-target, terms undisclosed. No candidate from it has ever been named. | Company-stated |
| Papers and patents | PubMed returned nothing attributable to the company. Patsnap records one item, a methods paper credited to Ankit Gupta, and zero medical patents. No granted US patent assigned to Reverie was found. | Not found |
| Federal funding | No NIH or NSF award under the company name. The platform was built on private capital alone. | Public record |
| Sale, as announced | Ginkgo's Feb 28, 2024 release describes acquiring “key assets of Reverie Labs”: infrastructure and software for training large scale AI foundation models, plus four AI staff including CTO Ankit Gupta. | Company release |
| Sale, as filed | Ginkgo's Q1 2024 Form 10-Q records 1,832,172 Class A shares issued to Reverie Labs, Inc. on Feb 23, 2024, worth about $2.4 million, for “certain infrastructure and software assets”. In the same filing Patch Biosciences was acquired by merger for about $14.4 million. Reverie's was an asset purchase; the drug programs appear in neither. | Public record |
| Retention economics | Ginkgo approved restricted stock units with an aggregate grant date fair value of $3,530,000 to four former Reverie employees, vesting over four years. More than the stock paid to the company. | Public record |
| Current status | reverielabs.com no longer returns a page; Internet Archive captures since December 2024 record redirects and 403 responses. Tracxn lists it as acquired, four employees as of August 2026. | Independent |
Read plainly: Reverie built real chemistry and a real platform and left a thin record of it. It published almost nothing, patented almost nothing, and never registered a trial. What outlived it is software, and the price of that software is the one hard number in the file. The drug candidates were not in the sale and have not been heard from since.
What to watch
- Whether RVL-101 surfaces again. If the asset moved, the first public sign would be a trial registration.
- Whether Roche or Genentech ever names a molecule from the 2020 collaboration. Six years on, none has.
- What Reverie Labs, Inc. does with 1,832,172 Ginkgo shares. Ginkgo announced a restructuring in May 2024 cutting labor expense by at least 25 percent.
- Whether the benchmark gap Gupta named gets filled. He argued there are no good public benchmarks for the tasks that actually bottleneck drug design.
- The wider scoreboard. A 2024 Drug Discovery Today analysis found “AI-discovered molecules have an 80-90% success rate” in Phase 1 but that “in Phase II the success rate is approximately 40%, albeit on a limited sample size”, in line with the industry it set out to beat.
In their words
“We began as a machine learning driven software company selling SaaS tools and consulting services to pharma companies, and at acquisition we were a pharmaceutical company, developing our own in-house pipeline of drug assets and advancing them rapidly using our machine learning technology.”
Ankit Gupta, co-founder and CTO, blog, January 2025 · Founder
“Suddenly, the TAM looks more like double digit millions instead of billions.”
Ankit Gupta, on sizing the pharma software market, 2025 · Founder
“it became clear that most biotech investors valued these programs at ~$0 beyond the reputation signaling that we weren’t total clowns”
Ankit Gupta, on how investors priced the milestones, 2025 · Founder
“It is a privilege to begin a long-term partnership with Roche’s and Genentech’s successful R&D organizations to develop life-changing medicines with the potential to transform patient care.”
Jonah Kallenbach, co-founder and CEO, July 2020 · Company release
“Ginkgo is a perfect home for the technology that Ankit, the Reverie team, and I have built over the last six years, and I am very excited to see the impact it will have in the short and long term.”
Jonah Kallenbach, on the acquisition, February 2024 · Company release
“Data generation in biology and drug development has accelerated dramatically over the last two decades, and the talent and skills Jonah, Ankit, and their team possess in machine learning and AI position them well to exploit that data to design better molecules to meet severe unmet needs.”
Wayne Holman, Ridgeback Capital, February 2021 · Investor, company release
“But if you know how these models work, it goes from 5% to maybe 6% or 7%.”
Patrick Malone, KdT Ventures, Endpoints News, 2023 · Independent
Related companies
Sources
- Public recordGinkgo Bioworks Form 10-Q, Q1 2024, Item 2
- Public recordClinicalTrials.gov search, Reverie Labs, nil return
- Public recordEDGAR search, no Form D filed by Reverie Labs
- Public recordFederal records pull, nil across FDA, NIH and NSF
- IndependentReverie Labs uses new machine learning algorithms
- IndependentA virtual drug discovery company with no lab of its own
- IndependentBioworks company forges 3 acquisition deals
- IndependentThe first AI-designed drugs fall short in the clinic
- IndependentAI-discovered drugs in clinical trials, 2026 survey
- IndependentReverie Labs, Inc. pipelines and patents
- IndependentReverie Labs company profile and acquisition
- FounderHow unfair is the coin?
- CompanyMulti-target collaboration agreement with Roche
- CompanyReverie Labs raises $25M Series A round
- CompanyGinkgo Bioworks acquires Reverie Labs platform
- CompanyGinkgo compensation information on recent acquisitions
- CompanyReverie Labs pipeline page
Profile researched and written by Healthcare Discovery. Last updated September 29, 2026.
