Probabl is a new AI company built around popular library scikit-learn

Probabl isn’t your average AI startup, as this new French company is an Inria spin-off company that revolves around an open source data science library called scikit-learn — Inria is a well-known French technology research institute.

As for scikit-learn, with more than 45,000 stars on GitHub, this Python module is widely used by machine learning teams working on tabular data. It can be used for model fitting, predicting, cross-validation, etc.

Unless you’re an ML developer, this might be the first time you’re hearing about scikit-learn. But many big companies have relied on the library for their own products, such as Spotify, Hugging Face, Booking.com and Dataiku.

Some of the contributors behind scikit-learn and other Inria-backed open source libraries have decided to create a company to ensure that these projects remain actively developed and properly funded with revenue-making activities.

Researchers who are joining this project include Camille Troillard, Gaël Varoquaux, Olivier Grisel, François Goupil, Guillaume Lemaitre, Jérémie Du Boisberranger and Fabien Gandon.

Yann Lechelle, the former CEO of cloud hosting company Scaleway, spent four months as Entrepreneur in Residence at Inria working on this project. Following the spin-off from Inria, he will assume the role of CEO of Probabl.

“We are a software publisher, but our first commercial offerings will include professional services, training and certification, specifically around scikit-learn,” Lechelle told me. “Our scope will be broader to cover the complete data cycle from organization and cleaning to machine learning and MLOps.”

As a company, Probabl now has three types of shareholders:

Public shareholders, such as Inria’s investment subsidiary Inria Participations, the French government through its French Tech Souveraineté program;Private shareholders, such as Costanoa Ventures;And finally individual contributors.

They will all play a part in the governing body of the newly formed organization. Probabl is calling itself a company with an industrial and digital sovereignty mission. As a result, it wants to release truly open source projects, which hasn’t been the case lately in the AI industry.