Maximilien Le Cleï

Professional Experience

• Convinced management and succesfully led the company in applying modern SOTA to imitate D-BOX haptic artist tracks (generative modeling of 1D signals – chair actuators displacement – with audio+video conditional diffusion transformers | state space models)
• Strategized with leadership, given modeling successes, to create previously infeasible solution offerings predicted to significantly expand D-BOX's ~40M dollar revenue generating customer base, with generous newfound profit margins given AI-centricity
• Training billion parameter models on terabytes of modern D-BOX data. Running carefully-crafted, systematic scaling experiments over task complexity, training regimes, architectural choices, input modalities, model capacities & dataset sizes
• PhD research application: continual learning & out-of-distribution generalization, building on 20+ years of D-BOX craftsmanship

• Attempting to evolve beyong the AI paradigm: rawer human behaviour modeling through the expansion of random space exploration
• Human agentic & brain behaviour imitation (focus: continual learning & out-of-distribution generalization), GECCO 2024 publication

• Led the development of a strongly automated and highly collaborative machine learning (deep learning + evolutionary machine learning) training workspace to speed up and strengthen the experiments of present machine learning researchers
• Technical contribution: audio model alignment with fMRI data to improve generalization, published in Imaging Neuroscience 2025
• Technical advising: large language model (LLM) fine-tuning + visual language model (VLM) for fMRI brain data alignment
• Evolutionary machine learning for human video game behaviour cloning & modern reinforcement learning benchmarks
• + work leading to publications. 1st author: ALOE@ICLR+GECCO 2022 & GECCO 2023; Imaging Neuroscience 2x 2025 & 1x 2026

• Research and development of novel evolutionary machine learning algorithms for imitation and reinforcement learning
• Developed i) a highly scalable and distributed open-source general purpose evolutionary machine learning library ii) a novel high-level imitation learning paradigm and iii) extremely compute efficient artificial neural networks

• Early-stage application of modern deep reinforcement learning techniques to industrial embroidery problems
• Developed embroidery virtual environments of various complexity/fidelity & applied state-of-the-art deep reinforcement learning algorithms (DQN, Prioritized experience replay & Dueling Double-Q-Learning) to prototype highly capable embroidering agents

Academic Publication

1. Y Harel, B Pinsard, JA Boyle, A Cyr, M Le Cleï, P Mignot, M St-Laurent, K Jerbi, P Bellec. Gamer in the scanner: Event-related analysis of fMRI activity during retro videogame play guided by automated annotations of game content. Imaging Neuroscience, 2026.
2. M Freteault, M Le Cleï, L Tetrel, L Bellec, N Farrugia. Alignment of auditory artificial networks with massive individual fMRI brain data leads to generalisable improvements in brain encoding and downstream tasks. Imaging Neuroscience, 2025.
3. A Kemtur, F Paugam, B Pinsard, Y Harel, P Sainath, M Le Cleï, J Boyle, K Jerbi, P Bellec. Behavioral imitation with artificial neural networks leads to personalized models of brain dynamics during videogame play. Imaging Neuroscience, 2025.
4. M Le Cleï, S Bar-Sheshet, P Bellec. Generational Information Transfer with Neuroevolution on Control Tasks. GECCO, 2024.
5. M Le Cleï, P Bellec. Generative Adversarial Neuroevolution for Control Behaviour Imitation. GECCO, 2023.
6. M Le Cleï, P Bellec. Neuroevolution of recurrent architectures on control tasks. ALOE@ICLR + GECCO, 2022.

Education

Software