Reinforcement Learning Researcher (Humanoid)

Reinforcement Learning Researcher (Humanoid)

Location: San Francisco, CA (On-site at REK HQ)
Reports to: CTO
Company: REK Inc.

About REK

REK is pioneering a new global sport: Robot Kombat. Its VR-controlled humanoid robot fighting.

We merge robotics, gaming, and live entertainment into the most advanced real-time human-machine competition on Earth. Our debut event, REK0, was the most attended robot fight in Temple Nightclubs history.

As we scale globally, were building a world-class robotics and AI team to push the limits of humanoid motion, balance, and control through reinforcement learning and sim-to-real transfer.

The Role

Were seeking a Humanoid Robot Reinforcement Learning Researcher to develop and train advanced control policies for our REK humanoid robots.

Youll work closely with REKs robotics, simulation, and teleoperation teams to design reinforcement learning pipelines that improve movement quality, stability, responsiveness, and adaptability all optimized for real-time fighting performance.

This role sits at the intersection of research and applied engineering: youll design algorithms, implement simulation environments, train agents, and test them on full-scale humanoid robots.

Responsibilities

• RL Algorithm Development Design and implement state-of-the-art reinforcement learning algorithms for humanoid locomotion, balance, and reactive movement.
• Simulation Environments Build and customize physics-based simulation environments (Isaac Gym, MuJoCo, PyBullet, etc.) for efficient training and domain randomization.
• Sim-to-Real Transfer Develop robust transfer strategies that ensure trained policies perform reliably on physical humanoid robots.
• Policy Evaluation Define performance metrics, run experiments, and benchmark results for both simulated and real-world tests.
• Integration & Collaboration Work closely with teleoperation and control system teams to blend RL policies with operator input in hybrid control architectures.
• Research & Publication Stay current with cutting-edge humanoid and robotics RL research and contribute to internal whitepapers or external publications as appropriate.
• Data & Infrastructure Maintain scalable training pipelines and data logging systems using GPU clusters or cloud resources.

Required Qualifications

• 2+ years of hands-on experience in reinforcement learning for humanoid robots at a university lab, research institute, or corporation.
• Deep understanding of deep RL algorithms (PPO, SAC, TD3, DDPG, etc.).
• Experience using Isaac Gym, MuJoCo, PyBullet, or Gazebo for simulation training.
• Strong software engineering skills in Python, PyTorch, and C++.
• Understanding of robot kinematics, dynamics, and control systems.
• Ability to run large-scale experiments efficiently and interpret quantitative results.
• Strong communication skills and comfort working in an experimental, cross-disciplinary team.

Preferred / Bonus Qualifications

• Experience with Unitree G1 EDU
• Background in teleoperation, imitation learning, or motion retargeting.
• Familiarity with low-latency communication protocols and embedded robot control.
• Understanding of VR systems, motion capture, or human-in-the-loop RL.
• Mandarin proficiency a major plus for collaboration with Chinese robotics partners and manufacturers.

Compensation

• Competitive salary based on experience.
• Equity participation in a rapidly growing robotics entertainment startup.
• Full benefits and access to REKs humanoid robotics lab and training infrastructure.

Why REK

At REK, your research will come to life on stage powering real humanoid robots in front of live audiences.
You wont just train agents in simulation youll see your policies executed in full-scale, physical combat, shaping the future of sport, robotics, and human-machine interaction.