BigHat Biosciences Builds an AI-Native Antibody Engine

BigHat Biosciences is building an AI-native system for designing protein therapeutics, with an unusual insistence that the software answer to the laboratory. Its platform connects frontier models to autonomous, high-throughput experiments, then feeds the resulting biological data back into the next design cycle.

The San Mateo biotechnology company was founded in 2019 by Peyton Greenside and Mark DePristo. Greenside now serves as CEO, while DePristo remains a co-founder and adviser. BigHat develops wholly owned medicines and works with pharmaceutical partners on difficult antibody-engineering programs.

BigHat matters now because its thesis has crossed into the clinic. Lead program BHB810 entered a Phase 1 study in 2026, a second candidate is moving through preclinical development, and a new $75M Series C brought company-reported funding to $223M. The question is no longer whether AI can propose an interesting protein. The question is whether a company can repeatedly turn designs into drug candidates with the manufacturing, safety, and clinical properties that medicine demands.

How BigHat's AI and wet lab work together

BigHat's core advantage is supposed to be a learning loop, not a single model. Milliner is the company's automated experimental engine. It produces and characterizes protein variants, generating the positive and negative results that models need to improve. Reccy acts as the data-control layer, while Reccy Antibody Design Studio, or RADS, coordinates models and datasets across programs.

That architecture attacks a recurring weakness in AI drug discovery: a model can generate more molecules than a traditional laboratory can reliably build, test, and learn from. BigHat designed the laboratory around the computational cycle, with standardized assays, automation, and software linking predicted properties to measured outcomes.

For partners, BigHat reports one- to two-week data-generation cycles, support across major antibody formats, and more than 20 assay types. Those figures are company-reported, but they reveal the operating ambition. BigHat is trying to make biological data production behave less like a series of bespoke experiments and more like an engineered system that improves with use.

The pipeline is becoming the platform's real test

BHB810 is BigHat's lead clinical program, a CDH17-directed antibody-drug conjugate for gastric cancer and other advanced gastrointestinal malignancies. BigHat announced the first patient dosed in September 2026 in a Phase 1 dose-escalation study. The company has reported encouraging preclinical findings, but no human clinical outcome data is available.

The distinction matters. AI-designed medicine has spent years collecting elegant demos, promising papers, and patient investors. Clinical development is less impressed by novelty. A molecule has to show acceptable safety, measurable activity, manufacturability, and a benefit that survives contact with human biology.

BHB299, an avidity-driven CEACAM6xCD3 T-cell engager for solid tumors, is BigHat's second named program. The company says it is nearing the end of preclinical development and plans to begin human studies in 2027. Additional oncology and immunology programs remain earlier or undisclosed.

Partnerships add a second route to validation

BigHat's partnering strategy gives the platform another way to prove itself while its internal pipeline matures. The company has announced collaborations with Amgen, Merck, Johnson & Johnson, AbbVie, and Eli Lilly across oncology, immunology, neuroscience, and broader protein-engineering challenges.

The most useful signal is completed work, not logo density. BigHat reported successful completion of a strategic J&J collaboration and three Merck project collaborations. Merck is continuing to evaluate sequences generated through those programs, and BigHat received an undisclosed milestone payment after the three projects.

This hybrid model is demanding. Partner programs can generate data, revenue, and external validation, but they also compete for scientific attention with wholly owned assets. BigHat has to show that the same infrastructure can serve different molecular problems without becoming a custom-services shop wearing an AI badge.

Leadership reflects a company moving into development

BigHat's current leadership spans computation, laboratory systems, drug development, and business operations. Peyton Greenside leads the company as CEO and co-founder. Stefan Weigand, formerly head of large-molecule research at Roche, is Chief Scientific Officer. John Corbin is Chief Development Officer, and Liz Schwarzbach is Chief Business Officer.

The operating bench also includes Eddie Abrams as Chief Information Officer, Charbel Eid as VP of Platform, Hunter Elliott as VP of Machine Learning, Ryan Henrici as VP of Discovery Medicine, and Brett Weinstein as VP of Finance. That mix tracks the company's central burden: BigHat has to run an AI organization, an automated laboratory, a drug-discovery engine, and a clinical-stage biotech at the same time.

Leadership changes also deserve precision. Mark DePristo helped found BigHat and served as its earlier CEO. Greenside is the current CEO. Treating those titles as interchangeable would make the story cleaner and the facts worse, a trade biotech has already made often enough.

Hiring shows where BigHat expects pressure

BigHat's careers page listed 21 open roles on September 24, 2026. The openings spanned laboratory science, machine learning, software engineering, clinical operations, finance, and corporate systems. A time-stamped job count can change quickly, but the functional spread says more than the total.

The company is staffing both sides of the loop. It needs people who can automate experiments, characterize molecules, build models, maintain data infrastructure, and move candidates through clinical development. That is a sign of execution load, not proof of commercial success.

BigHat describes its culture as high-performance, collaborative, family-friendly, and patient-focused, with an emphasis on integrity and work-life balance. The real cultural test will be whether teams with radically different working rhythms can share one operating system. Machine-learning researchers iterate in code. Clinical organizations advance through documentation, controls, and decisions that cannot be hot-fixed on Friday night.

What BigHat Biosciences must prove next

The September 2026 $75M Series C gives BigHat more runway to advance BHB810, prepare BHB299 for the clinic, expand its data platform, and support partnered work. It also raises the standard of proof.

BigHat now has to deliver clinical readouts without confusing preclinical promise for patient benefit. It has to show that its data advantage compounds across programs, that partner successes translate into repeatable economics, and that a larger organization can preserve the speed of its design-build-test-learn loop.

The company has built an impressive machine for asking biology better questions. The next chapter is about the answers: whether BigHat can turn AI, automation, and disciplined experimentation into medicines that work for patients, not just models that look smart on a slide.