Jesse Read

This is a Hi!Paris Advanced Synergy Fellowship (internal chair), between LTCI Télécom Paris (co-PI: Albert Bifet) and LIX Ecole Polytechnque, co-financed by BNP Paribas.

We will be hiring a PhD *and postdoctoral researcher on this topic for early 2027.

This project addresses the challenge of distribution shift (or concept drift) in real-world AI systems, where changing data conditions can make deployed models unreliable, unpredictable, or even unsafe. We propose a unified framework that formulates robustness to drift (whether natural or adversarial, gradual or abrupt) as a predictive maintenance problem for autonomous systems. The project will focus on agent-driven model-agnostic interventions; promoting a reduce, reuse, and recycle methodology — to enable robust maintenance even for opaque or proprietary systems. By combining uncertainty-aware planning with model-based inference, and with simulation-based reasoning, we aim to establish the foundations of a new class of Frugal, Agent-based, Robust Methodology for Streams (FARMS).