About Us
Our robots generate massive multi-modal data streams, from video, audio, proprioception, to control trajectories. To learn from this at scale, we build the simulation and data infrastructure that turns real-world and virtual experiences into structured data for embodied agents.
This role sits at the core of that system, creating the environments and data that power large-scale robot learning.
Role Overview
You will architect and maintain the simulations & data platform powering our robot learning stack, ensuring high-fidelity data capture, scalable synthetic data generation, and seamless real2sim/sim2real integration.
Responsibilities
- Develop and maintain simulation environments for embodied learning and reinforcement learning: using tools such as Isaac Sim, Mujoco/MJX
- Generate synthetic data at scale for vision, audio, and control tasks: integrate with model training pipelines (e.g., PyTorch, JAX, Ray, or RLlib)
- Design multimodal data systems: structuring streams from sensors, cameras, IMUs, and actuators into training-ready datasets
- Implement real2sim/sim2real adaptation techniques: domain randomization, latent-space alignment
- Collaborate on policy learning loops: connecting simulation rollouts to real-world deployment for continuous improvement
- Contribute to engine and infrastructure tooling: from low-level C++ optimization to high-level Python interfaces
Preferred Qualifications
- C++ and/or Python proficiency
- Experience with physics engines or robotics simulation software
- Deep interest or familiarity with Robotics simulators: Isaac Sim, Coppelia, Genesis, Bullet, or RaiSim, etc.
- Deep interest or familiarity with physics engines: MuJuCo/MJX, PhysX, Pybullet, or Havok, etc.
- Deep interest or familiarity with tools: ROS, Moveit, etc
- Bonus: Strong background in physics, graphics, or math
- Bonus: Practical experience implementing Vision-Language Action Models in robotic systems
Bonus Skills
- Built or contributed to robotic simulations or data systems.
- Experience with foundation model data curation (tokenization, sharding, filtering).
- Strong interest in enabling embodied AI through scalable data infrastructure.
Top Skills
What We Do
Menlo Research is an open AI & Robots lab.
We build the brains for robots. It’s time to tell robots what to do!
