Head of Reinforcement Learning

Position details

•         Department: Reinforcement Learning Department, Foundation AI Division, VinRobotics

•         Reports to: Director, Foundation AI Division

Purpose of the role

Lead the Reinforcement Learning Department in building RL-based controllers for VinRobotics humanoids: locomotion, whole-body control and loco-manipulation, from simulation to real robots. The role owns the RL training and sim-to-real pipeline, taking policies onto the VR-H3, VR-M3 and VR-H5-D platforms and into real use cases.

Key responsibilities

1.       Technical strategy: build and execute the RL roadmap for locomotion, whole-body control and loco-manipulation; work with the Robot Foundation AI department so low-level policies take commands from VLAs or high-level planners, and with the Edge AI Optimization department so policies run in real time on onboard hardware.

2.       Locomotion and whole-body control: perceptive and terrain-aware locomotion, payload carrying, whole-body motion tracking from mocap and egocentric data, compliance and force-aware control for heavy humanoids.

3.       Training and sim-to-real pipeline: GPU-parallel simulation (Isaac Lab, MuJoCo/MJX, mjlab, Genesis), domain randomization, teacher-student distillation, system identification and actuator modeling; evaluation and safety procedures for hardware testing.

4.       Field deployment: meet stability, robustness and safety targets for policies in factory use cases and demos; work with the mechanical, electronics and firmware teams through product QCD.

5.       Research and IP: lead new directions (behavior foundation models, cross-embodiment control, RL fine-tuning of VLAs, model-based RL); publications, patents and open-source releases; technical collaboration with partners and universities.

6.       Team leadership: hire to headcount, run the KPI process, and mentor engineers and talent-program members (interns, Young Engineers, Residents).

Requirements

•         MSc or PhD in Computer Science, Robotics, Control Engineering, Mechatronics or a related field.

•         5+ years in RL or robot control with policies running on real robots, including 2+ years leading an engineering team.

•         Solid command of RL algorithms (PPO, SAC, off-policy, model-based), imitation learning, reward design, curricula and sim-to-real.

•         Strong Python and PyTorch or JAX; hands-on with GPU-parallel simulators (Isaac Lab, MuJoCo/MJX, Genesis).

•         Deep understanding of robot dynamics and control (kinematics, dynamics, PD, whole-body control, MPC).

•         Professional working English for direct collaboration with international partners.

Preferred

•         Has deployed locomotion or whole-body control policies on real humanoid or legged robots.

•         Experience with motion retargeting and tracking from mocap/video, behavior foundation models, or RL fine-tuning of VLAs.

•         C++, ROS 2, real-time systems and onboard policy deployment.

•         Publications at top venues (CoRL, RSS, ICRA, IROS, NeurIPS, ICLR) or impactful open-source contributions.