At Jolibrain, I design and develop critical AI-based solutions, from reliable deep learning training and inference pipelines to multimodal LLM specialization software.
I've delivered work on anomaly detection, embedded deep architectures, vision-based landing, metrology, software quality, and graph neural network heuristics for SAT solving with organizations including Airbus, CNES, SNCF, Thales, Vinci, and Kratos.
I develop reinforcement learning algorithms based on graph neural networks for decision-making under uncertainty, with applications in earth observation, factory and supply-chain scheduling, and aerial path planning.
Previously at ONERA and the University of Maryland, I worked on planning, Markov decision processes, constraint programming, robot task planning, model checking, and spatio-temporal databases. I also contribute through academic research, teaching, training, consulting, and published research.
