Robotics Reinforcement Learning Engineer

Focus: Humanoid Control/Loco-Manipulation/ Dexterous Hands/Whole-Body Control

Responsibilities

1. Research and develop reinforcement learning (RL) algorithms for robotic skill acquisition, including robotic arm manipulation, dexterous hand control, and whole-body control.
2. Explore the integration of reinforcement learning with end-to-end robot manipulation policies and Vision-Language-Action (VLA) models, such as ACT, Diffusion Policy, OpenVLA, and π0.
3. Design and implement RL training pipelines for robotic manipulation using simulation platforms such as Isaac Gym, Isaac Sim, Isaac Lab, and MuJoCo, with a focus on effective sim-to-real transfer.
4. Develop real-world reinforcement learning methods using human feedback, model-based feedback, and online interaction to improve policy generalization, precision, and robustness.
5. Combine imitation learning and reinforcement learning to enable efficient acquisition and generalization of complex robotic manipulation skills.
6. Optimize, distill, and deploy learned policies on physical robots, ensuring reliable, stable, and efficient real-world performance.
7. Stay at the forefront of robotics and reinforcement learning research, translating cutting-edge advances into practical algorithms and robotic capabilities.

Qualifications

1. Master's degree or PhD in Robotics, Computer Science, Artificial Intelligence, Control Engineering, or a related field.
2. Strong programming skills and hands-on experience with deep learning frameworks such as PyTorch, TensorFlow, or JAX.
3. Solid understanding of reinforcement learning fundamentals, with practical experience in algorithms such as PPO, SAC, TD3, or DDPG.
4. Hands-on experience with robotics simulation platforms such as Isaac Gym, Isaac Sim, Isaac Lab, or MuJoCo, including sim-to-real transfer.
5. Good understanding of robot kinematics and dynamics, with experience in robotic manipulation, motion planning, or control.
6. Familiarity with end-to-end robotic manipulation policies, including ACT, Diffusion Policy, OpenVLA, π0, or other VLA models.
7. Experience in reinforcement learning or imitation learning for robotic arms, dexterous hands, legged robots, or humanoid robots is highly preferred.
8. Experience with distributed training, large-scale model training, or the development of RL training infrastructure is a plus.
9. Publications at top-tier conferences in robotics or machine learning (e.g., RSS, CoRL, ICRA, IROS, NeurIPS, or ICML), or strong performance in relevant robotics competitions, are a plus.
10. Strong technical curiosity, a passion for robotics, and excellent problem-solving skills, with the ability to independently explore and implement new ideas.