Ayoub Ajarra

Hello,

I am Ayoub, a third year PhD candidate at Inria in Scool team (previously SequeL) under the supervision of Debabrota Basu and Philippe Preux.

My research interests are broad. Prior to my PhD, I was a visiting research student at MILA, supervised by Yoshua Bengio, Loubna Benabbou and Dianbo Liu, where I was working on enabling AI models to learn physics with no feedback in the context of a physicist (Navier-Stokes equations) by relaxing “physical” inductive bias, and how we can rigorously reduce uncertainty in chaotic (turbulent) regimes to learn space-time evolution of a dynamical system.

My research spans learning theory, algorithmic fairness, multi-armed bandits, reinforcement learning, curriculum learning and sequential decision-making. In my PhD work, I develop frameworks that enable policymakers to detect and assess algorithmic bias in realistic settings, particularly, in dynamic and competitive markets where companies continuously adapt their systems throughout the audit process. I’m also deeply drawn to the mathematical beauty of statistical learning and the surprising connections they reveal across combinatorics, topology, and statistics.