Stanford's AI Lab Partner Helps 15,000 Scientists Faster

Scientists racing to cure diseases just got a powerful new teammate that never sleeps, never gets overwhelmed by data, and can juggle thousands of tasks at once.

Stanford researchers built Biomni, an AI assistant that works like a virtual lab partner for biomedical scientists. Instead of drowning in spreadsheets and spending weeks analyzing mountains of data, researchers can now ask Biomni to handle the heavy lifting while they focus on the creative breakthroughs.

The platform combines advanced language models with over 150 specialized research tools, 59 scientific databases, and 100 software packages. Scientists simply type what they need into a chat window, like "analyze this glucose data and find meaningful patterns," and Biomni creates a plan, writes code, and gets to work.

"Today, we have abundant biomedical data, but we don't have enough human researchers to analyze it all," says Kexin Huang, one of Biomni's creators and former Stanford PhD candidate. The AI can integrate knowledge across multiple scientific fields and manage tasks that would take human teams months to complete.

In one real example, a researcher gave Biomni 458 Excel files tracking glucose monitors from 30 people over several months. The AI autonomously created a 10-step analysis, identified meal patterns from glucose spikes, extracted temperature readings, and delivered a clear report highlighting both individual details and broader trends.

The Ripple Effect

Since becoming open source, Biomni has attracted more than 15,000 scientists who've automated 100,000 different research workflows. They're using it to formulate hypotheses, perform complex genetic analyses, and design rigorous experiments across fields from metabolic research to precision medicine.

The platform excelled in initial testing on established biomedical benchmarks and performed well on eight challenging real-world scenarios it had never seen before. While it still struggles with tasks requiring deep clinical judgment or novel experimental reasoning, it already matches human-level performance in database querying, sequence analysis, and molecular cloning.

Every conversation between scientist and AI gets fully documented, so researchers never lose their work history in scattered notebooks. Scientists can monitor each step and jump in anytime to adjust the direction.

The success led the team to launch a startup called Phylo in September 2025, with venture funding and an Academic Lab Program for universities. The original code remains fully open source so researchers worldwide can build on it.

"Biomni is a real partner in biomedical research," says Stanford computer science professor Jure Leskovec, who co-created the platform. The team envisions a future where virtual AI biologists routinely work alongside humans to dramatically speed up discoveries that save lives.

For thousands of scientists grinding through tedious data analysis, that future just arrived.