Locomotion Researcher (AI & Robotics Lab)

Hyphen Partners; Singapore; On-site; SGD 120,000 - 150,000; Full time

Job summary

Hyphen Partners is partnering with an Applied R&D lab focused on open-source humanoid robot platforms. They seek a Locomotion Researcher to develop motion capabilities from walking to running with real hardware.

You will own the sim2real pipeline, push robustness, design reward shaping in MuJoCo, and work across hardware, controls, and manipulation researchers. You will open-source what you learn.

Qualifications

• Deep hands-on experience with reinforcement learning for continuous control, legged locomotion, or whole-body control.
• Track record of getting learned policies onto real robots.
• Strong command of a physics simulator such as MuJoCo or Isaac.
• Fluency in Python and modern RL tooling, and comfort in a ROS2-based control stack.
• Bias for shipping: prefer real hardware progress this week over perfect in simulation.
• Clear thinking about why a policy fails, not just whether it does.

Responsibilities

• Design, train, and ship reinforcement learning policies for bipedal locomotion and whole-body control.
• Own the sim2real pipeline end to end, building on zero-shot sim2real work.
• Push balance, gait, and recovery beyond the demo and into real-world robustness.
• Build and refine reward design, domain randomization, and training environments in MuJoCo.
• Close the loop between simulation and hardware with real telemetry from robots.
• Collaborate with hardware, controls, and manipulation researchers.
• Open-source what you can and document learnings for the community.

Skills

Reinforcement learning for continuous

Legged locomotion

Real-robot policy deployment

Python

ROS2

Tools

MuJoCo

Isaac

ROS2

Job description

Our client is an Applied R&D lab that focuses on building open-source humanoid robot platforms, and the full software stack that powers it. Their mission is to make humanoid labor economically viable, turning software into physical labor at scale, building across the full stack: hardware architecture, locomotion, autonomy, simulation, and infrastructure.

The Role

They are looking for a Locomotion Researcher to develop motion capabilities from walking to running, recovering, and moving through the real world with the confidence of something alive. You will own the policies that make that motion robust on real hardware, and will not stop at a clip in simulation. You will get your work onto a physical biped and keep pushing until it holds up under contact, disturbance, and terrain that they did not train for.

What You’ll Do

• Design, train, and ship reinforcement learning policies for bipedal and whole-body locomotion
• Own the sim2real pipeline end to end, building on their zero-shot sim2real work so a first run on hardware is never really a first run
• Push balance, gait, and recovery behavior past the demo stage into something that survives pushes, slips, and uneven ground
• Build and refine reward design, domain randomization, and training environments in MuJoCo
• Close the loop between simulation and hardware with real telemetry from real robots, then feed what breaks back into the next policy
• Work shoulder to shoulder with hardware, controls, and manipulation researchers, since whole-body control does not respect team boundaries
• Open-source what you can and write up what you learn so the community can build on it

What We Look For

• Deep hands-on experience with reinforcement learning for continuous control, legged locomotion, or whole-body control
• A track record of getting learned policies onto real robots, not only into papers or simulators
• Strong command of a physics simulator such as MuJoCo, Isaac, or similar, including reward shaping and domain randomization
• Fluency in Python and modern RL tooling, and comfort in a ROS2-based control stack
• A bias for shipping: you would rather see a policy stumble on real hardware this week than look perfect in sim next month
• Clear thinking about why a policy fails, not just whether it does

Nice to Have

• Published work in locomotion, legged robotics, or sim2real transfer
• Experience with model predictive control or classical locomotion methods alongside learning-based approaches
• Contributions to open-source robotics or RL projects
• Experience bringing up new hardware and debugging the messy gap between a model and a motor