Titouan Guerin
PhD student in deep reinforcement learning, physics-informed models and model-based control at ONERA, the French Aerospace Lab, from November 2026.
about
I studied AI and Machine Learning at Sorbonne Université (the MIND Master's, previously called DAC), specialising in probabilistic modelling, deep learning and reinforcement learning. Before that I completed a Bachelor's in Computer Science at the University of Luxembourg in 2023.
I am drawn to the mathematics behind machine learning, and to reinforcement learning in particular. My final Master's internship at ISIR and ONERA was on RL and physics-informed models for drone control, and in November 2026 I am continuing that work there as a PhD. Outside research I have done some software development and a little tutoring.
things i know my way around
- rl
- model-free · model-based
- deep learning
- computational graphs · autodiff · cnns · gnns
- llms
- transformers · pre-training · adaptation · evaluation · rag
- generative
- vaes · gans · flow matching · diffusion
- ai4science
- physics-informed models · neural odes · world models
- vision
- cnns · vision transformers · self-supervised learning · vlms · vlas
- and
- mlops · ai ethics: logic programming · causal reasoning · responsibility
Most of my M2 coursework is on GitHub, in M2-MIND: homeworks, notes and small projects. Maybe it can help future students (it does not include most of the big projects, those I kept for myself ;)
phd
Model-based deep reinforcement learning for drone control with physics-informed priors
From November 2026 I am a PhD student at ONERA in Palaiseau, funded by the ANR, working on model-based deep reinforcement learning for drone control with physics-informed priors.
The thesis is co-directed by Julien Marzat (Université Paris-Saclay) and Olivier Sigaud (Sorbonne Université), and supervised by Pierre Fournier at ONERA.
It continues my Master's work on fixed-wing UAVs: what a physical prior brings to the learned dynamics model, then the harder problems around it, long horizons, sparse rewards, partial observability and generalising across tasks and aircraft.
selected work
publications
Beyond Chatbots: Enhancing Luxembourgish Language Learning Through Multi-agent Systems and Large Language Model
Nouzri, S., El Fatimi, M., Guerin, T., Othmane, M., Najjar, A.
PRIMA 2024: Principles and Practice of Multi-Agent Systems, Kyoto. Lecture Notes in Computer Science, vol. 15395, Springer · 2025
education & experience
- 2026.11
- PhD student, ONERA - Université Paris-Saclay & Sorbonne Université. Model-based reinforcement learning for drone control.
- 2026.03
- Research internship, ISIR x ONERA. World models and deep reinforcement learning for drone control.
- 2024.07
- AI research assistant, AI Robolab, University of Luxembourg. Backend for a multi-agent language learning app built with LangGraph. Co-authored a paper presented at PRIMA in Kyoto.
- 2023.09
- MSc AI and Machine Learning, Sorbonne Université
- 2023.02
- Information security intern, Grant Thornton Luxembourg
- 2022.08
- Data analysis intern, Grant Thornton Luxembourg
- 2022.02
- Semester project tutor, University of Luxembourg
- 2020.09
- BSc Computer Science, University of Luxembourg
- 2020.09
- Augmented reality intern, Goodyear. HoloLens 2 maintenance guide.
familiar tools
- ml
- python · pytorch · gymnasium · scikit-learn · tensorboard · tensorflow · torchrl · numpy · pandas
- web
- typescript · git · fastapi · react · next.js
- systems
- sql · bash · docker · linux