Daphne Koller

CEO & Founder — insitro

Profile

Daphne Koller is one of the foundational figures in modern machine learning — she spent two decades at Stanford helping make probabilistic graphical models a workable tool for real problems, and she co-wrote the textbook that taught everyone else how to do it. Before deep learning ate the world, PGMs were how you reasoned under uncertainty at scale, and Koller’s framing of Bayesian networks, Markov random fields, and structured inference shaped how a generation of researchers thought about machine learning. She was a MacArthur Fellow in 2004, is a member of the National Academy of Engineering and the National Academy of Sciences, and her work sits underneath a lot of the probabilistic intuitions that still show up in modern systems.

Developers mostly know her, though, for what she did next. In 2012 she and Andrew Ng turned Stanford’s experiment in free online CS courses into Coursera, and for a brief window it felt like universities might actually be disrupted. That didn’t quite happen, but Coursera did genuinely democratize access to AI education — if you learned ML from a MOOC in the 2010s, you probably learned it from a course she helped make possible. Her 2012 TED talk on online education is still the clearest articulation of what MOOCs were trying to be.

In 2018 she left Coursera to found insitro, a drug-discovery company betting that ML trained on massive, purpose-generated biological datasets can find drug targets humans can’t see. The thesis is important: insitro doesn’t just bolt models onto existing pharma pipelines — they generate their own data (induced pluripotent stem cells, high-content imaging, functional genomics) specifically so the models have something real to learn from. In January 2026 they acquired CombinAbleAI and launched TherML, a full-stack ML platform spanning small molecules, oligonucleotides, and antibodies, with their first drug candidate heading into clinical trials this year.

For anyone learning AI, Koller is worth studying for a specific reason: she’s the rare researcher who’s operated successfully in three completely different modes — deep theory, mass education, and applied industry — without losing rigor in any of them. The through-line is that she keeps choosing problems where ML could matter and then actually doing the work.