Stagiaire Recherche en Intelligence Artificielle / Apprentissage par Renforcement (F/H)

Valeo is a tech global company, designing breakthrough solutions to reinvent the mobility. We are recruiting a Research Trainee in Artificial Intelligence, specifically in reinforcement learning. The proposed internship is structured around participation in the Waymo Open Sim Agents Challenge (WOSAC), an international competition that establishes the state of the art in behavioral simulation.

Requirements

• Master the PufferDrive simulator and implement the evaluation pipeline and metrics of the WOSAC challenge.
• Implement reference models (Baselines): Train and evaluate initial agents based on Imitation Learning (Behavior Cloning) and Simple Reinforcement Learning (PPO) to establish a baseline score.
• Innovate and optimize algorithms: Design advanced architectures (Transformers, Graph Neural Networks) and explore Multi-Agent Learning (MARL) to improve the realism and diversity of simulated behaviors.
• Analyze and value results: Perform robustness tests (closed-loop simulation), write a scientific report, and potentially submit an article to an international conference (e.g., NeurIPS Workshop, ICRA).

Benefits

• Presence of a Social and Economic Committee (CSE): Travel, social and cultural activities
• Transportation reimbursement up to 70%