I am a postdoctoral researcher at CREST (ENSAE Paris, Institut Polytechnique de Paris), working with Arnak Dalalyan. My research has two strands. The first is the mathematics of generative models: how diffusion and flow matching models behave in the small-noise regime, and how their samplers propagate errors. The second is human–AI collaboration in mathematical research. I designed and built Human–AI Mathematics, which maps the tools for mathematics with AI and develops conjecture search, an open framework in which AI agents search for and cross-review proofs, and every claim they certify can be checked by a mathematician.
I came to this from both research and industry. After a PhD in statistics at École Polytechnique on Langevin sampling, and an internship at Google Brain, I spent six years building machine learning systems: signal processing, object detection and anomaly detection at Thales, deep learning for drug discovery at Iktos, then foundation models for biology at MBZUAI and GenBio AI. In 2025 I worked closer to end users, building proofs of concept for AI decision-support tools and freelancing on LLM systems, from evaluating LLM answers at Rollstack to improving retrieval at Clarifeye.
What ties it together is taking an idea from paper to working system, and measuring whether it actually works. Publications are on the Research page, code on Projects, and longer notes on the blog. To get in touch, send me an email.
My postdoctoral work is funded by the European Research Council (ERC) under the European Union’s Horizon Europe research and innovation programme (grant agreement No. 101201229).
