I am a Research Scientist in the Optimization Department of EDF Lab Paris-Saclay, where I develop and refine advanced numerical solvers tailored for energy management applications. Our team tailors diverse optimization methodologies based on problem structure, with solutions encompassing mixed-integer, continuous, stochastic, distributed optimization, and machine learning methods.

Previously, I was a PhD student at Inria and ENS Paris, specializing in numerical optimization and machine learning applied to real-time robotics. My PhD research was conducted within the Willow and Sierra teams and under the supervision of Justin Carpentier, Adrien Taylor and Jean Ponce. During my PhD, I notably developed the open-source ProxSuite library, which integrates the ProxQP quadratic programming solver, and the QPLayer differentiable layer.

Prior to my PhD, I obtained MSc degrees in Applied Mathematics at the École Polytechnique and in Statistical Learning at the University of Cambridge (tripos part III). As a French senior civil servant of the "Corps des Ponts, des Eaux et Forêts", I also hold a master of public administration from the École des Ponts et Chaussées.
Updates:

  • November 2024: Our contribution to the CORL 2024 workshop on Differentiable Optimization has been accepted!

  • October 2024: I have given a talk to the Center of Automation and Control of Mines Paris Tech on two classical problems of Centralized Energy Production Management. We provide context, current algorithms used, and some enhancements we have worked on. Checkout the talk page if interested.

  • June 2024: Our contribution to the RSS 2024 workshop on frontiers of optimization has been accepted! Checkout our here.

  • September 2024: I joined EDF Lab as a Research Scientist! I will be working on numerical optimization for centralized energy management.

  • March 2024: I participated to the 2024 ICLR conference ! I presented a spotlight poster about differentiating over infeasible QPs at ICLR conference. Check out as well our paper here.

  • January 2024: I defended my PhD! You can find the slides on the Talks page.

  • December 2023: We submitted a a new journal paper . It extends a line of work on more numerically robust differential dynamic programming for generic trajectory optimization (using proximal methods) to which I had the pleasure to participate.

  • September 2023: Check out our new preprint (submitted at journal Transaction and Robotics) proposing a few enhancements to ProxQP solver (in terms of numerical robustness or speed, or extension to non convex settings) along with real time robotic evaluations. We propose as well in the form of a companion report a global convergence proof as well.

  • June 2023: My submission for ICLR 2024 has been accepted as a spotlight paper ! It proposes techniques for differentiating over infeasible QPs. Check out the paper here here.

  • May 2023: My poster for ICRA 2023 has been accepted (presenting new techniques for learning new types of quadratic programs leveraging augmented lagrangian methods and primal infeasibility). Check out the talk page for more info.

  • May 2023: Our contribution for differentiating over collisions has been accepted at ICRA 2023. Check out the publication page for more info.

  • November 2022: I had the pleasure to contribute to the design of more robust differential dynamic programming using proximal methods. Our contribution has been accepted at IROS 2022 ! Check out the publication page for more info.

  • July 2022: I am participating at RSS 2022 conference! See out our ProxQP presentation at 6th paper session.

  • July 2022: The ProxSuite library to be presented at RSS 2022 is officially released !

  • May 2022: I had the pleasure to contribute to an extension of ProxQP for nonlinear optimization on manifold. This contribution was presented at the 6th edition of the Workshop on Legged Robots (ICRA 2022). Check out the paper for more info.

  • March 2022: Our RSS submission presenting a new Quadratic Programming solver (coined ProxQP) for real time robotics has obtained been accepted! The paper is available here.