Reinforcement Learning Engineer (Whole Body Control)

Join our team as a Reinforcement Learning Engineer, where you will develop, train, deploy, and evaluate advanced reinforcement learning algorithms for whole body control of our humanoid robot. You will determine the observations, actions, and model types that unlock maximum performance, identify and close sim-to-real gaps, and define, test, and evaluate performance metrics for learned policies. You should have experience tuning hyperparameters and cost functions for RL algorithms, familiarity with common RL techniques, and a strong background in dynamics and control, ideally of legged robots.

• Développer, entraîner et déployer des algorithmes d'apprentissage par renforcement pour le contrôle du corps entier.
• Déterminer les observations, les actions et les types de modèles qui maximisent les performances.
• Identifier et combler les lacunes les plus importantes entre la simulation et la réalité.
• Experience tuning hyperparameters and cost functions for these RL algorithms
• Capable of leading complex controls projects and mentoring junior engineers
• Familiarity with common RL techniques such as: domain randomization, curriculum learning, reward shaping, etc
• Experience with reinforcement learning algorithms for robotics: PPO, SAC, etc
• Strong background in dynamics and control, ideally of legged robots
• Experience with behavior cloning techniques (e.g. distillation)