About me

I’m a Machine Learning Engineer at Meta, where I work on integrity and trust — building systems that help keep platforms safe at scale.

Before London, I spent 10 years in Paris, where I studied at École Polytechnique then completed my Ph.D. in Applied Mathematics at Université Paris-Cité under Stéphane Gaïffas and Emmanuel Bacry. My research sat at the intersection of statistical learning theory and practical ML: I co-developed WildWood, a random forest algorithm that achieves gradient-boosting-level performance while staying interpretable and lightweight, published in IEEE Transactions on Information Theory. I also worked on deep learning for healthcare, modelling long-term disease relapses from claims data (ZiMM, Journal of Biomedical Informatics). You can find my doctoral thesis and defense slides.

These days, beyond my work at Meta, I’m increasingly drawn to the tooling layer of AI — how we build, ship, and interact with ML systems. I spend a lot of time thinking about developer experience, agentic workflows, and what the next generation of AI-native tools looks like in practice.

I’m based in London with my family. You can find me on GitHub, LinkedIn, or Google Scholar.

Publications

WildWood: a new Random Forest algorithm. IEEE Transactions on Information Theory. Joint work with I. Merad, S. Gaïffas. arXiv:2109.08010 · code

ZiMM: a deep learning model for long term and blurry relapses with non-clinical claims data. Journal of Biomedical Informatics, 2020. Joint work with A. Kabeshova, B. Lukacs, E. Bacry, S. Gaïffas. arXiv:1911.05346 · Short version at Machine Learning for Health Workshop, NeurIPS 2019 (ML4H)