Startups, Artificial Intelligence

Decentralized research platform Owkin raises another $18M for series A

Owkin, a startup building a decentralized health research platform, rounded out its series A round with another $18 million in funding. It uses federated learning to allow researchers to collaborate on AI models while keeping data onsite.

A startup building a platform for decentralized research rounded out its series A round with another $18 million in capital.

Owkin, which has its headquarters in New York and Paris, is building a research network using federated learning, a type of machine learning that allows researchers to test AI models on distributed data. Instead of having to move data to a centralized location, the model can be trained on data at disparate locations, such as multiple research labs or pharmaceutical companies. Some technologists have touted this as a way to maintain privacy and security while developing models.

“The idea of the federated learning technology is to allow researchers to collaborate on project without having to aggregate data into a consortium or a specific setting,” said Anna Huyghues-Despointes, Owkin’s head of strategy. “Our plan is to deploy this at scale.”

Mubadala Capital and Bpifrance led the most recent round of funding; the latter also led a $25 million raise announced last month.  Alphabet’s GV was also an early backer of the company.

Owkin has raised a total of $70 million for its series A since 2018. Huyghues-Despointes explained the lengthy funding round:

“We knew we had to raise significant capital going forward to invest in this,” she said, adding that through the different extensions, the company had been able to show more technology, research and revenue proof points.

The company was started in 2016 by Thomas Clozel, a former oncologist, and Gilles Wainrib, a computer science researcher who was a professor at the Ecole Normale Supérieure in Paris. In those early days, federated learning was quite new, and seldom used outside of academic research. But medical centers and pharmaceutical companies have warmed to the idea.

The startup is currently coordinating two consortia: MELLODDY, a collaborative drug discovery project between 10 pharmaceutical companies, including Amgen, Bayer and Novartis; and HealthChain, a collaborative project between four medical centers.

Owkin has also developed its own network of hospitals and research institutions to allow researchers to train models on real-world data to improve treatments or expedite drug development.  The company offers its technology to academic research centers for free. It has paid agreements with seven of the top 15 pharmaceutical companies, where it offers translational research and drug development services.

If you can’t see the data, how do you know if it’s any good? Owkin gets a de-identified sample of the data before starting a project. As the AI model is trained on the data, it can also monitor it for variance and other metrics for accuracy.

“There are a lot of methodologies to ensure quality of the data is high and training is meaningful,” Huyghues-Despointes said. “Our models are not black boxes. We put a lot of effort into understanding why we’re getting that prediction.”

The company’s goal is to expand to the top 30 research sites in the U.S. and the top 50 in Europe, as well as satellite locations in Asia and South America.

“Our mission is to advance medical research to be able to deliver better treatments for patients,” Huyghues-Despointes said. “Nothing we do is just for the sake of AI and machine learning. It’s really for collaborative research and science.”

Photo credit: Gremlin, Getty Images

Shares0
Shares0