Paper Club #7 — Gaël Varoquaux, Probabl

A one-hour paper session on tabular foundation models, the family of models that has started displacing gradient boosting and random forests on tabular data over the past two years. Gaël Varoquaux is speaking. He co-created scikit-learn, is Chief Science Officer at Probabl and Research Director at INRIA, and his team builds open-source tabular foundation models including TabICL and TabICLv2. He covers how the models actually work, what the recent research says, and where he thinks tabular machine learning goes next. Run by Unaite, the student federation across France's main AI institutions, at Parisanté Campus in the 15th. Registration is approval-gated. Why go: a rare chance to hear the scikit-learn co-creator on the one corner of ML where deep learning has not obviously won. Best for data scientists and ML engineers who work with tabular data every day.