Medra raises $52M Series A to build autonomous AI scientists that run lab experiments at scale
Feb 23, 2026 with Michelle Lee
Key Points
- Medra raises $52M Series A led by Human Capital to build autonomous labs that run drug discovery experiments at scale, opening a fully autonomous facility in San Francisco in 2026.
- The company positions itself as a data foundry for biology, generating clean experimental data at scale to train foundation models that currently have three orders of magnitude less training data than large language models.
- Medra's differentiator is physical AI that reasons in real time during experiments, not just mechanical precision, allowing its robots to adapt mid-experiment like human scientists would.
Summary
Michelle Lee founded Medra to combine life sciences, robotics, and AI. The company builds autonomous systems that run laboratory experiments at scale for drug discovery, positioning itself as infrastructure rather than a direct drugmaker.
Lee studied chemical engineering and earned a PhD in robotics at Stanford under Jeannette Vojillo and Fei-Fei Li, where she built robotics foundation models. She started Medra to apply that work to early-stage drug discovery.
Medra's robots use off-the-shelf hardware with proprietary AI layered on top. The differentiator is not mechanical precision—traditional lab automation can pipette consistently—but physical AI autonomy that reasons in real time about experiments. Lee describes it as capturing what expert scientists do when they read papers, sense reactions in the lab, smell conditions, visualize results, and adjust mid-experiment. Medra's vision language-action models attempt to encode that flexibility and scientific reasoning.
The business model works two ways. Medra partners with pharma and biotech firms such as Genentech, who can run experiments in-house using Medra's systems or send work to Medra's own lab. The Series A funds the opening of a fully autonomous lab in San Francisco in 2026, which Lee describes as one of the largest autonomous labs in the United States.
Lee calls Medra a "data foundry" for life sciences, comparing it to TSMC in chip manufacturing. Biology foundation models train on three orders of magnitude less data than large language models like O1. Clean, at-scale biological and chemical data does not exist on GitHub or the open internet and must be generated. Medra's autonomous lab is built to produce that data generation infrastructure, allowing pharma partners to train foundation models in biology.
Medra raised $52M in Series A funding led by Human Capital, with participation from returning investors Lux Capital and Menlo Ventures, plus Cataglio. The round funds scaling data generation capacity and opening the San Francisco lab, not custom hardware or foundation model training.