Decentralized Asynchronous Optimization for Deep Learning.Supervised by Edouard Oyallon and Eugene Belilovsky.

• (Willow team) Weakly supervised semantic segmentation with Vision Transformers.

(Cognitive Machine Learning team) Unsupervised learning of speech representations. Supervised by Emmanuel Dupoux.

• Study of the computational complexity of a multilevel combinatorial problem over graphs• Design of heuristics using Reinforcement Learning and Graph Neural Network to solve the problem• Redaction of two research papers : one in computational complexity, the other in Machine Learning [NeurIPS 2020, Journal of Computer and System Sciences 2022]Research done under the supervison of prof. Margarida Carvalho

• Created a non-profit association which main purpose was to promote music to disadvantaged children through daily workshops• Lead the 14 members of the team and managed the project during its 2 years preparation and a 7 months trip around the world• Gave concerts around the world, supervised the production of a CD and the making of 2 films about the project

• Developed pricing tools for the automobile insurance• Built a new model to predict the risk of water damage