Paul Mangold is an Assistant Professor at École polytechnique, where he works within the SIMPAS team. He is also a Hi! PARIS Fellow for the 2026-2029 period.

His research focuses on optimization, federated learning, reinforcement learning, and the ethical challenges related to training machine learning models. He is particularly interested in designing algorithms that adapt to the structure of problems in order to improve convergence, communication efficiency, privacy, and fairness.

Before joining École polytechnique as an Assistant Professor in 2025, he was a postdoctoral researcher there under the supervision of Éric Moulines. He completed his PhD at Inria Lille, within the Magnet team, with a thesis on privacy-preserving optimization and machine learning. He previously studied at ENS de Lyon and completed the Master Datasciences at Université Paris-Saclay. His broader interests also include uncertainty quantification, learning with missing data, and developing practical and trustworthy machine learning methods.