Inductive Bio Sweeps Three ADMET Challenges, Wins $21M Grant

Three Wins, $21 Million, and What They Buy

A Series A drug‑discovery AI company has spent the past year doing something rare in a field crowded with big‑lab challengers: beating them, repeatedly, on the blind benchmarks they helped design. Inductive Bio, a New York startup built by former Flatiron Health machine‑learning leaders Josh Haimson and Ben Birnbaum, racked up three consecutive first‑place finishes across the industry's flagship ADMET prediction competitions between early‑2025 and mid‑2026, then converted that record into an up‑to‑$21 million federal award aimed at replacing animal testing with AI safety models. The wins and the award arrived weeks apart and gave the company something money alone could not buy: a verified, third‑party ranking it can put in front of skeptical pharma procurement teams.

The competition track began in the inaugural Polaris ADMET challenge, where Inductive's Beacon‑1 model placed first among 39 competitors from leading AI drug‑discovery groups and academic labs. The dataset was real: a recently disclosed coronavirus main protease program, with models scored blind on chemistry the teams had never seen. That win coincided with the company's $25 million Series A in May 2025, led by Obvious Ventures with Andreessen Horowitz Bio + Health, Lux Capital, S32, Character, and Amino Collective participating. Obvious partner Rohan Ganesh framed the bet plainly: Inductive had "moved beyond the hype of AI in drug discovery to deliver measurable results."

Seven months later, the bigger test came. In the OpenADMET‑ExpansionRx Blind Challenge, built around a real myotonic dystrophy, ALS, and dementia drug‑discovery campaign and run with Expansion Therapeutics, Inductive's Beacon models placed first among more than 370 submissions from large pharma, biotech, academic labs, and AI organizations. The challenge asked entrants to predict nine critical ADMET endpoints covering absorption, distribution, metabolism, excretion, and toxicity. "We're proud to place first among competitors from some of the largest companies in the space, demonstrating for a second time that Inductive's Beacon models represent the state of the art in ADMET prediction," Haimson said at the time.

In July 2026, Inductive made it three in a row, taking the OpenADMET PXR Blind Challenge against more than 350 researchers. The task was narrow: predict activation of the pregnane X receptor, a protein that flags foreign compounds for metabolism and a liability that frequently surfaces only after years of lead optimization. Across the three blind challenges, more than 750 competitors made nearly 10,000 submissions spanning 17 critical assay endpoints, and Inductive placed first in all three, beating entrants from Merck (with NVIDIA, Novo Nordisk, and EMD Serono), outfits with research budgets orders of magnitude larger. "Three consecutive wins in independent blind challenges against teams 1000x our size tell us that the Beacon models powering that lab are the state of the art for AI drug discovery tasks like ADMET prediction," Haimson said.

Then came the federal money. In December 2025, ARPA‑H awarded Inductive up to $21 million under the Computational ADME‑Tox and Physiology Analysis for Safer Therapeutics (CATALYST) program to lead a project called DATAMAP, or Digital Acceleration of Toxicity Assessment with Mechanistic and AI‑driven Predictions. The team (Inductive, Amgen, Cincinnati Children's Hospital Medical Center, Baylor College of Medicine, and Torch Bio, a University of Michigan Medical School spinout) will generate biological data using organoids, ex‑vivo human tissue, and microphysiological systems, then train AI models on it. The first targets are drug‑induced liver injury and cardiotoxicity, which together drive nearly 40% of post‑market drug withdrawals. Inductive will work with the FDA on potential regulatory use cases; Amgen will advise on real‑world context and, if the technology holds, submit an investigational new drug application built on the in‑silico safety package. Birnbaum, principal investigator on the award, pointed to the underlying failure rate — roughly 90% of clinical‑stage drug candidates never reach the broader market, and a quarter of those failures trace to safety issues preclinical animal testing missed — to argue for human‑biology‑based predictive models.

The combination gives Inductive two distinct currencies at once. The challenge wins prove its models work on chemistry no one at the company has seen, validated against competitors with far deeper pockets. The ARPA‑H award proves the U.S. government is willing to bet taxpayer dollars on the same approach, with a built‑in FDA pathway. Together, they form the credibility stack the company is now leaning on to staff its first dedicated enterprise sales function.

What the Partnerships Listing Really Signals

In the wake of the OpenADMET‑ExpansionRx blind challenge win and the up‑to‑$21 million ARPA‑H CATALYST award, Inductive posted a Partnerships Lead role on its careers page: based in New York City, San Francisco, or Boston, paying $225,000 to $275,000 plus equity. The timing is the point. The listing functions less like a routine backfill than like a declaration that the Series A company's go‑to‑market function is being formalized to convert scientific credibility into signed enterprise contracts.

The role's scope reflects a deliberate shift in how Inductive intends to monetize the Compass platform that, by May 2025, already supported dozens of active small‑molecule programs and had logged more than one million molecule designs explored by medicinal chemists. Until now, Compass adoption has grown largely through technical champions inside discovery teams. The Partnerships Lead opening suggests the company now wants a dedicated owner for the relationship layer above that, someone who can take an OpenADMET win or an ARPA‑H contract and turn the resulting visibility into multi‑year commitments from large pharma.

Several features of the listing support that reading. The band tops out at $275,000, materially above what Inductive has historically paid for pure research hires and within the range where enterprise sales and business‑development leaders in AI drug discovery typically sit. The three‑city geography tracks the metros where pharma procurement, biotech business development, and AI‑diligence functions concentrate, rather than the single‑site footprint a research scientist would require. "Hybrid" implies a field‑facing remit that includes on‑site work with customers.

The mandate also dovetails with context the company has already established. Inductive is a co‑launcher of ADME‑One with Ginkgo Datapoints and Tangible Scientific, a high‑throughput ADME platform designed to move Absorption, Distribution, Metabolism, and Excretion decisions earlier in drug discovery. That kind of multi‑party commercial arrangement requires someone whose job is to manage partnership economics, not just scientific deliverables. And the ARPA‑H award, which runs through a multi‑institutional team including Baylor, Torch Bio, Amgen, and Cincinnati Children's, creates a federally‑funded channel into Amgen that a Partnerships hire can broaden into adjacent commercial programs once the cooperative‑agreement work is done.

In short, the listing reads as the first explicit hire of a go‑to‑market enterprise function inside a company that previously relied on founder‑led selling and inbound technical interest. Whether it is a single role or the start of a small team is not yet public, but its existence marks the moment Inductive stopped treating its deal pipeline as a byproduct of its science and started treating it as a function in its own right.

You can view the live Inductive Bio Partnerships listing on Zero G Talent for current status and comp details.

What the Wins Are Already Buying

The commercial signal behind Inductive Bio's go‑to‑market push is unusually concrete for a Series A company at this stage. The clearest of those signals is the ADME‑One™ joint platform Inductive launched with Ginkgo Bioworks Datapoints and Tangible Scientific, announced in May 2026 on top of a strategic partnership the three companies had signed in August 2025. ADME‑One pulls all five Tier 1 assays (microsomal stability, cell permeability, kinetic solubility, CYP inhibition, and plasma protein binding) into a single end‑to‑end workflow run inside Ginkgo's automated Boston lab, then layers Inductive's AI for human pharmacokinetic projection and Tangible's compound management. The pitch is explicit: move ADME profiling out of late lead optimization and into hit identification, where "derisking is most needed," and return results in days rather than weeks at a price positioned below offshore CRO alternatives — a deliberate play for U.S. and European teams reshoring preclinical work under the BIOSECURE Act. To run product strategy on the Ginkgo side of that joint platform, Ginkgo Datapoints hired Jonathan Grob, a Novartis and Valo Health veteran, as VP of Small Molecules, confirmation that the alliance is being operated as a real product line.

The ARPA‑H CATALYST award doubles as an enterprise pipeline asset. The up‑to‑$21M DATAMAP project pairs Inductive with Amgen, Cincinnati Children's, Baylor, and Torch Bio to build AI toxicity models for those two endpoints, which drive nearly 40% of post‑market drug withdrawals. The program's most consequential commercial clause is the requirement that Amgen "demonstrate these technologies' real‑world impact by submitting an investigational new drug application (IND) to the FDA using the technology." That is a procurement signal as much as a research deliverable: Amgen is committing to put the resulting models on a path that ends in a regulatory filing, not a paper.

The third indicator is Inductive's late‑June 2026 entry into Anthropic's Connector ecosystem for life sciences, which surfaces its ADMET prediction capabilities directly inside Claude for medicinal chemists. The integration targets what Haimson has called the "last mile problem": powerful models that working chemists never reach. Combined with Inductive's placement atop three consecutive OpenADMET‑style blind challenges and its reported delivery of multiple development candidates in timelines "significantly shorter than the industry standard," the enterprise story has stopped looking like promise and started looking like throughput.

One open thread: the company has not publicly named most of the "leading biopharma partners" it now counts. Denali Therapeutics has been a public partner since 2023; the rest of the roster remains undisclosed, which is the kind of detail a more mature GTM function usually starts pulling into the open.

How the Field Is Responding

Recursion arrived at this fight with the deepest ammunition cache. By the close of 2025, the Salt Lake City TechBio reported a pipeline spanning 10 clinical and preclinical programs and over 10 advanced discovery programs across oncology and rare diseases, roughly 580 employees ("Recursionauts"), and what its 2024 annual report calls "the most advanced supercomputer in the pharmaceutical industry," BioHive‑2. Its enterprise book matters most for this comparison: Recursion has banked $213 million in upfront and milestone payments from its Roche and Genentech partnership, with six Phenomaps accepted and one small‑molecule program initiated, plus a separate $134 million from a second pharma collaboration. The company is also closing in on clinical validation of its full AI operating system through REC‑4881 in familial adenomatous polyposis, where 75% of evaluable patients showed reductions in total polyp burden and a 43% median reduction after 12 weeks — numbers the company framed as proof that AI‑driven biological insight can translate to patient outcomes.

Recursion's countermove against Inductive's ADMET franchise is less a head‑on assault than a pivot toward owning the full stack. Its 2024 annual report leans hard into the argument that AI is commoditizing and that dataset scale is the moat: "We believe that rapid commoditization of artificial intelligence creates a once‑in‑a‑generation market opportunity for companies with the ability to build the right datasets in biology and chemistry to win." Recursion's own preclinical metrics (REC‑617 hitting candidate in under 12 months on 136 novel molecules, advanced candidates averaging ~330 compounds per program in ~17 months versus an industry baseline of more than 2,500 compounds and 42 months) are pitched as evidence that the integrated platform beats point solutions on cycle time. The argument, stripped down, is that pharma will pay for the discovery engine that surrounds the ADMET problem, not a better ADMET model alone.

Insitro is making a different adjustment. The South San Francisco company operates an AI-driven drug discovery platform and has, per Zero G Talent's job board, an active leadership bench with compensation bands that span roughly $224,000 to $480,000 across roles from Senior Director, Translational Medicine and Diagnostics to Chief Medical Officer, with a board‑wide median around $242,000 across 16 salaried postings, selective, not a blitz into enterprise sales.

The wider field is fragmenting into two responses. On one flank, platform‑first incumbents like Recursion argue that ADMET is a feature, not a company. On the other, consortium‑and‑distribution plays bundle Inductive's validated models with wet‑lab capacity: ADME‑One packages Inductive's predictions with high‑throughput Tier‑1 assays and integrated compound management at a price several times below the industry standard, pitched explicitly against offshore CROs and described as a fully U.S.‑based workflow. Each move widens the surface area where Inductive's benchmark wins turn into pharma touchpoints, even as the Recursions of the field insist the right answer is still the full AI operating system.

Defending Share Against Big Tech

The same blind challenges that gave Inductive a podium finish are also the place where it is most visibly outgunned on paper. In the most recent OpenADMET PXR competition, the company placed first overall but beat out teams fielded by Merck (working with NVIDIA) and those other two, groups, as the company put it, that came from "companies many times its size." STAT's reporting on the same result sharpened the asymmetry: a 28‑way statistical tie among the top finishers suggested the real advantage sat in training data quality rather than model architecture, and that any well‑resourced incumbent with comparable data access could close the gap.

That risk is structural. Lilly has put $1 billion into an NVIDIA‑backed AI lab; Isomorphic Labs and Recursion are pushing AI‑designed molecules into Phase I; Isomorphic is signing fresh 2025–2026 deals with major pharma. Atomwise has evaluation deals with Pfizer. The "Big Pharma builds its own" playbook is well funded and actively executing. For a Series A company with $29.4M raised to date, the durable counter is the asset the giants cannot replicate overnight: a pre‑competitive data consortium that now spans enough partners that a single member leaving would not collapse the flywheel, plus the validation record from winning three consecutive blind benchmarks.

Inductive is making three moves to widen that moat. First, the Anthropic partnership makes its ADMET models available as a Claude connector, embedding the company's predictions inside the workflow tools pharma scientists already use, which raises switching costs for any in‑house team that might otherwise build a substitute. Second, the ARPA‑H award ties Inductive to Amgen, Baylor, Cincinnati Children's, and the FDA on liver‑ and cardiotoxicity models — the two largest drivers of clinical drug safety failures. If those models clear regulatory scrutiny, the company will own a category of validated, regulator‑accepted predictions that no Big Tech entrant can shortcut. Third, the ADME‑One joint platform, priced below the industry standard and pitched specifically against offshore CROs in the post‑BIOSECURE environment, gives Inductive a U.S.‑based lab‑in‑the‑loop revenue stream that compounds data into the consortium every time a new campaign runs.

The harder problems are commercial, not technical. Inductive reported in mid‑2025 that it was a roughly 10‑person team hiring to 12, with ambitions to add pharma partners and academic groups to the consortium at a pace that depends on closing scientists who could otherwise land at Recursion or insitro. The partnerships lead role Inductive is currently advertising is the front line of that fight: every senior account hire who carries pharma relationships becomes a moat against the Big Tech land grab, and every quarter without one is a quarter incumbents use to sign multi‑year in‑house agreements.

The realistic 12‑month test is whether Inductive can convert the next 12 pharma pilots into multi‑program consortia memberships before Merck‑with‑NVIDIA, Isomorphic, or an in‑house Big Pharma team fields the next entry that ties or beats it. Beacon can keep winning benchmarks and the gap will still close. What decides who owns ADMET prediction as an industry standard is the data consortium, the regulatory pathway, and the enterprise sales bench, and Inductive is, right now, hiring the last of those three.

Working in frontier tech? Zero G Talent tracks the openings: see every open Insitro role, browse frontier tech jobs, the companies hiring, and the people building the field.