This page lists my scientific publications and ongoing work. Most of my research focuses on embedding physical priors into neural network architectures for interactive physical simulation in virtual reality. They stem from my work at CEA-List, with the exception of the NeurIPS workshop (EDF research internship) and the NLP publication (freelance mission).
Video reconstructed from a latent space trajectory generated by a Lagrangian Neural Network, for a 2D (top) and 1D (bottom) latent space.
Preprint, 2026. arXiv: 2608.16324.
Latent space trajectory obtained by sweeping a 360° view of an office. Our method (right) produces a smooth trajectory without discontinuities.
Learned latent kinetic and potential energy landscapes for a 2-DOF nonlinear dynamical system.
7th International Workshop on Model Order Reduction Techniques (MORTech 2025).
Predicted deformed configuration (left) and latent strain energy (right) for a 20-DOF hyperelastic beam.
Journal of the Mechanics and Physics of Solids, Vol. 194, January 2025, 105953.
Example of causality graphs and the associated DGLSTM network employed to extract information from ST code.
Predicted and simulated fluid velocity fields, along with the error field, for the von Kármán test case using our GNN architecture.
Key result: our architecture structurally prevents the emergence of stable orbits.