Thomas Clozel was a hospital doctor when he decided the future of medicine would not be decided on the ward. He was an assistant professor of clinical haematology at Henri Mondor, in the Paris suburb of Creteil, surrounded by patients and by data that mostly stayed locked in filing systems.
He came to medicine with an unusual family background. His parents, Jean-Paul and Martine Clozel, built the Swiss drug company Actelion and later started Idorsia. So he grew up near a model of how a scientist becomes an entrepreneur. He chose the clinic first. He trained in Paris and New York, and he has said that as a doctor he saw how advanced technology was becoming, while the benefits reached patients slowly.
Hospitals cannot simply hand over identifiable patient records, and privacy rules make that sensible.
The itch was specific. Drug development, he argues, is not working well, and biology is too complex to be reasoned out by humans alone. In 2016 a friend introduced him to Gilles Wainrib, a researcher with a doctorate in applying artificial intelligence to biology. Clozel supplied the clinical eye, Wainrib the machine learning. They founded Owkin in Paris that August.
The obstacle was built into their chosen field. Hospitals cannot simply hand over identifiable patient records, and privacy rules make that sensible. The company leaned on federated learning, where the model travels to the data rather than the other way round. In finance and industry that was established. In healthcare it was close to unused.
What Clozel remembers best is not a funding round. In an interview he recalled the moment the first models analysing tissue slides began to work, and later a model called MesoNET that predicted survival in mesothelioma patients better than the classification pathologists use. That, he said, convinced him the approach could work.
The money followed. In November 2021 Sanofi put in 180 million dollars, valuing Owkin at more than a billion. Swiss operations opened near Basel from October 2021, at the Idorsia site his parents had founded, and by 2022 two of its cancer diagnostic tools had European certification. In 2024 the Paris hospital network AP-HP renewed its partnership for at least five years.
The takeaway is plain. A doctor's frustration, paired with someone who can build the tool, is a stronger starting point than a market chart. The friend who made the introduction turned out to matter as much as the investor who wrote the big cheque.