Resume

Visiting research student - Supervised by Yoshua Bengio and Loubna Benabbou

MILA, Université de Montréal

Studied the ability of ML models to predict space time evolution of a dynamical system (fluids) with almost no knowledge of a physicist. Developped a new approach based on Helmholtz theorem to reduce the representation -describing Rayleigh-Bénard instability, into a discrete one and study the interaction of the vortices represented by keypoints, then implemented a graph encoder-decoder to learn the space-time evolution of the discrete representation based on causal graphs.

• Space-time evolution,SImulation, Deep learning for scientific discovery, Physical inductive bias, Causal graphs

PhD student

Scool (Previously Sequel) Inria - Université de Lille

Learning Theory, Fairness, Multi-armed Bandits, Reinforcement Learning theory and Sequential Decision Making