Robotics Learning Engineer
Descrizione dell'offerta
Responsible for developing and deploying learning-based control strategies for humanoid robot models, with a focus on reinforcement and imitation learning. The role centers on building robust training pipelines, designing diverse task environments, and managing experimentation workflows from simulation to real-world validation. It involves translating learning algorithms into reliable robot behaviors, addressing sim-to-real challenges, and ensuring performance through systematic benchmarking and analysis. Close collaboration with control, simulation, and hardware teams is required to align learned policies with physical system constraints and operational goals.
Responsibilities
● Design, implement and iterate RL / IL training pipelines for humanoid tasks in simulation and in the real world.
● Design and implement diverse task suites for manipulation, navigation, and whole-body coordination in simulation.
● Manage training experiments and evaluation loops: Hyperparameter tuning, Benchmarking, Logging and Failure analysis.
● Collaborate closely with multidisciplinary teams including: Motion and control engineers, Mechanical design team, Simulation engineers.
● Support sim-to-real transfer by adapting policies to real robot constraints and validating performance during deployment.
Requirements
● Ms or Phd in Robotics, Automation, Computer Science or related field.
● At least 2 years experience in ML / RL / robotics.
● Strong Python + PyTorch. You can profile, debug numerics, and write maintainable code.
● Familiarity with RL algorithms (PPO, SAC, etc.) and robotics (states, control, kinematics).
● Experience solving real problems using reinforcement learning policies in any domain.
● Strong ownership mindset with ability to document experiments and communicate trade-offs clearly.
● Professional English level.
Nice to Have
● Experience with:
○ Isaac Lab / RL Games / Stable Baselines3
● Exposure to:
○ Imitation Learning / Behavior Cloning
○ Robotics datasets (state + vision)
● Experience with sim-2-real challenges