About me

I am an Applied AI Scientist at Mistral AI, working on LLM post-training, reinforcement learning, and agentic systems. I train and evaluate agents that solve complex tasks through tool use and interaction with their environment.

My research interests include:

• Reinforcement learning for LLMs: policy optimization, reward modeling, credit assignment, and curriculum design, with a focus on training stability and sample efficiency through analysis of importance sampling ratios, KL divergence, policy entropy, and gradient norms.
• On-policy distillation: distilling reasoning and tool-use capabilities into smaller, more efficient coding and research agents.
• Deep research and agentic systems: multi-step reasoning, tool use, information synthesis, and long-horizon task execution.

I am a graduate of École Polytechnique, where I completed the cursus d’ingénieur polytechnicien with a foundation in applied mathematics. I also hold a Ph.D. in Natural Language Processing and an M.S. in Data Science from Université Paris-Saclay.

I developed my mathematical problem-solving skills through university competitions, including the International Mathematics Competition for University Students (IMC 2012).