Thomas Revol

Biography

Thomas Revol is an Affiliated Student Researcher at ICSI, where he works with the AI and Big Data Research Group led by Dr. Michael Mahoney. 

Revol’s current research lies at the intersection of deep learning optimization and numerical linear algebra. At ICSI, he focuses on matrix-aware optimization, specifically working to extend the AutoSpec framework’s learnable recurrence family (introduced by Liu et al. in “Learning to Discover Iterative Spectral Algorithms“). His project aims to discover a broader class of iterative algorithms for matrix-function approximation by introducing nonlinear, higher-order matrix-state transitions and rational recurrences. While a major focus of this work is accelerating large-scale neural network training such as learned polar decomposition for Muon and inverse-root computation for Shampoo, the primary methodological contribution is highly versatile. By generating task-adapted executable algorithms from spectral descriptions, Revol seeks to apply these learned recurrences to diverse scientific computing challenges outside of AI, including sparse linear solvers, finite-element simulations, computational chemistry, and quantum physics.

Previously, Revol engineered PyTorch pipelines to extract and evaluate layer-wise embeddings from EEG Foundation Models, designing attention-based fusion mechanisms for clinical predictions. His diverse background also includes simulating collective animal behaviors using multi-agent modeling frameworks and engineering statistical pipelines for quantum state classification.

Beyond his research, Revol is a Franco-American dual citizen who is active in student leadership and athletics. He served as the Treasurer for the Télécom Paris Sports Organisation in 2025, is an active judo national competitor, and is training in triathlon.

Revol is pursuing a Master’s in Engineering at Télécom Paris (Institut Polytechnique de Paris) with a dual concentration in Data Science & Artificial Intelligence and Stochastic Modeling & Scientific Computing.

Publications

Areas of Expertise

• Machine Learning (Supervised, Unsupervised, Reinforcement Learning)
• Deep Learning and Neural Networks
• Generative AI and Foundation Models
• Multi-Agent Systems and Collective AI
• AI in Healthcare, Biomedicine, and Life Sciences
• Algorithms, Complexity Theory, and Computability
• Statistical Modeling, Bayesian Methods, and Inference
• Mathematical Optimization, Operations Research, and Decision Science
• Quantum Computing, Quantum Algorithms, and Cryptography

ICSI Activity