Robotics AI Engineer

Menlo Research; Singapore; On-site; SGD 120,000 - 190,000; Full time

Job summary

Menlo Research seeks a Robotics AI Engineer to design and train RL and imitation policies that run on the Asimov humanoid platform in the real world. You will work with hardware, firmware, and infrastructure teams to close the loop between training and deployment.

You will run experiments on physical robots, build fast data pipelines, and iterate policy improvements with systematic debugging and domain adaptation. This role requires hands-on work with real hardware.

Qualifications

• Strong foundations in reinforcement learning or imitation learning.
• Hands-on experience training policies that run on real systems.
• Proficiency in Python and familiarity with RL/ML frameworks (JAX, PyTorch, IsaacGym/IsaacLab, MuJoCo).
• An empirical, debugging-first mindset, caring about what actually works on hardware.
• Ability to move fast and context-switch between research problems and engineering tasks.

Responsibilities

• Design and train RL and imitation learning policies for locomotion, manipulation, or whole-body control.
• Run experiments on physical hardware and close the sim-to-real gap through systematic debugging and domain adaptation.
• Build and maintain simulation environments and data pipelines that support fast policy iteration.
• Instrument robot deployments and analyze failure modes to feed improvements back into training.
• Collaborate with hardware and firmware engineers to understand physical constraints and improve policy robustness.

Skills

Reinforcement learning

Imitation learning

Python

RL frameworks

Debugging mindset

Hardware experimentation

Context-switching

Tools

JAX

PyTorch

IsaacGym/IsaacLab

MuJoCo

Job description

About Menlo

Menlo Research is an Applied R&D lab building Asimov, an open-source humanoid robot platform, and the full software stack that powers it. Our mission is to make humanoid labor economically viable, turning software into physical labor at scale. We build across the full stack: hardware architecture, locomotion, autonomy, simulation, and infrastructure. We move fast, ship to real robots, and open-source everything we can. If you want your work to matter beyond a paper or a demo, this is the place.

The Role

As a Robotics AI Engineer, you will work at the intersection of learning and hardware, training and deploying policies that run on Asimov in the real world. This is not a research role in the traditional sense. You will be expected to get results on physical robots, not just in simulation, and to iterate fast when things break. You will work closely with the hardware, firmware, and infrastructure teams to close the loop between training and deployment.

What You Will Do

• Design and train RL and imitation learning policies for locomotion, manipulation, or whole-body control
• Run experiments on physical hardware and close the sim-to-real gap through systematic debugging and domain adaptation
• Build and maintain simulation environments and data pipelines that support fast policy iteration
• Instrument robot deployments and analyze failure modes to feed improvements back into training
• Collaborate with hardware and firmware engineers to understand physical constraints and improve policy robustness

What We Are Looking For

• Strong foundations in reinforcement learning or imitation learning, with hands-on experience training policies that run on real systems
• Comfort working directly with robots, not just simulators
• Proficiency in Python and familiarity with standard RL/ML frameworks (JAX, PyTorch, IsaacGym/IsaacLab, MuJoCo, or similar)
• An empirical, debugging-first mindset, you care about what actually works on hardware
• Ability to move fast and context-switch between research problems and engineering tasks

Nice to Have

• Prior work on humanoid or legged robot platforms
• Experience with sim-to-real transfer techniques (domain randomization, system identification, noise injection)
• Contributions to open-source robotics projects
• Background in control theory, trajectory optimization, or dynamics

Why Join Menlo

The policies you train do not sit in a notebook. They run on a real humanoid, in the real world, on short feedback loops. You will see your work move physical hardware within days, not quarters. You will collaborate directly with the hardware, firmware, and infrastructure teams, with high visibility and real stakes, and everything you can open-source, you will. If you want to build the systems that turn software into physical labor, this is where it happens.