Drug candidates frequently succeed in animal studies only to fail in human trials. Olio Labs believes the problem may lie not only with animal biology, but also with the narrow and outdated assays used to observe it.

Conventional animal studies measure a small number of predefined endpoints at fixed intervals. Adverse effects that emerge outside those windows can easily go undetected. Olio Labs is building an AI system that continuously records mouse behavior and physiology and links those signals to human clinical outcomes.
Founded by Systems Neuroscientists
Olio Labs was founded in 2023 by systems neuroscientists David Tingley and Tom Roseberry.
Tingley earned a PhD in neuroscience from New York University and completed postdoctoral research at Harvard. He has published more than 15 scientific papers, including work connecting oscillations in brain activity with glucose metabolism.
Roseberry has more than 15 years of experience across aerospace, brain-machine interfaces, synthetic biology and drug development. He previously co-led the development of a next-generation neural interface at Neuralink and later held leadership roles spanning engineering, computational biology and drug development at longevity biotech company Loyal. Tim Krausz, a co-author of the company’s technical article, is Olio’s director of engineering rather than a co-founder.
The Y Combinator S23 company has raised $5 million in total, according to CB Insights. Its latest disclosed financing was a $4.5 million round in March 2025, led by Boom Capital Ventures and Overwater Ventures. Pioneer Fund, SciFounders, Scrum Ventures, Y Combinator and ZAKA Ventures also participated.
Capturing Every Movement
Olio began with an algorithm designed to identify combinations of molecules that could improve efficacy while reducing side effects. During Y Combinator’s Summer 2023 program, the founders sent 16 obesity-drug combinations selected by the algorithm to a contract research organization in China.
The company says 12 combinations produced greater weight loss and lower nausea-related readouts than semaglutide. Each test, however, cost approximately $4,000 per combination and took about 60 days. The experimental process could not keep pace with an algorithm capable of generating trillions of candidate combinations.
Accuracy was an even larger concern. One combination that appeared successful in the outsourced assay caused visibly severe sickness when Olio tested it internally. The CRO had measured body weight, food intake and a brief daily health check, missing adverse behavior that occurred during the rest of the day.
Olio responded by building the Phenotyping Acquisition Device, or PAD. Cameras and sensors continuously capture movement, breathing, sleep and social behavior. Computer vision models quantify subtle changes such as turns, tremors and interactions between animals.
Rather than assigning a biological meaning to every behavior in advance, Olio asks which combinations of behavioral features predict measured outcomes in human trials. The company says the approach can predict clinical nausea rates and could eventually extend to gastrointestinal events, neurotoxicity, cardiac toxicity and patient-specific responses.
Turning In Vivo Studies Into a Data Platform
Olio has also built a customized laboratory information management system, QR-based sample tracking and vision-AI auditing. Its aim is to compare more than 20 candidates in parallel and allow AI agents to recommend subsequent experiments. The company expects to collect more than five million hours of mouse data by mid-2027.
The approach differs from biological AI systems trained primarily on sequences, structures or cellular images. Blank Bio focuses on RNA-level hypotheses, while Recursion applies machine learning to large-scale cellular imaging. Olio is targeting the translational gap between animal responses and human outcomes.
Olio has not formally called its system a foundation model. A more accurate description is a reusable, organism-level drug-response model designed to support multiple efficacy and safety predictions.
FDA’s Shift Away From Animal Testing
The US Food and Drug Administration has not banned animal testing in preclinical drug development. The FDA Modernization Act 2.0, passed in 2022, removed the blanket statutory requirement that every investigational drug be supported by animal-test data. It allowed sponsors to use alternatives such as cell-based assays and computational models when they can provide suitable evidence.
In April 2025, the FDA announced a roadmap to reduce animal testing in preclinical safety studies. The agency initially focused on monoclonal antibodies and selected drugs, encouraging organoids, organ-on-a-chip systems, human-cell assays and AI-based computational models known collectively as New Approach Methodologies, or NAMs.
The transition has since expanded through more specific guidance for certain monoclonal antibodies and oncology drugs. It remains a phased, product-specific effort to reduce unnecessary animal studies where validated alternatives exist—not an immediate prohibition covering every drug and animal experiment.
For Olio, the policy creates both risk and opportunity. Regulators and drug developers increasingly favor methods grounded directly in human biology. A platform that still conducts new mouse experiments is not, by itself, a non-animal alternative. If organoids, human-cell systems and computational models become sufficiently predictive, demand for conventional animal efficacy and toxicology studies could decline.
The opportunity lies in the transition period. Olio is not proposing a higher-volume animal-testing CRO. Its platform is designed to extract more information from each experiment, monitor animals continuously and reduce repeated studies. That could support the FDA’s principles of reducing, refining and ultimately replacing animal use.
Its longer-term value may also shift from the PAD hardware to the accumulated dataset and predictive models. If Olio can reliably learn the relationship between mouse behavior and clinical outcomes, some future predictions could potentially be made from historical data and computation without commissioning another animal study.
The decisive question is therefore not whether Olio can automate animal experiments. It is whether its data can predict human efficacy and adverse events better than conventional assays—and whether that performance can be reproduced across independent clinical trials and therapeutic areas.
MORE FROM THE POST
- Biomolecular AI Startup Boltz Raises $110M Series A Led by Bessemer
- AI Drug Design Lab Isomorphic Labs Raises $5B Series C
- a16z Says AI Will Make Pharma Even Bigger. Here’s What the Numbers Show
- Google DeepMind Unveils AlphaGenome 2 to Read How DNA Variants Act in Specific Cell Types
- Formation Bio Launches $600M Fund to Buy and Develop Stalled Clinical-Stage Drugs
Share
Most Read
- 1
- 2
- 3
- 4
- 5


Leave a Reply