AI Training Infrastructure Engineer (Robotics)

TL;DR

AI Training Infrastructure Engineer (Robotics): Designing and scaling the training and deployment backbone for RL-based whole-body control systems with an accent on simulation, data pipelines, and hardware deployment. Focus on optimizing cluster utilization, integrating physics engines, and accelerating iteration cycles for humanoid robots.

Location: San Jose, CA (Require 5 days/week in-office collaboration)

Salary: $200,000–$300,000 annually

Company

AI Robotics company developing autonomous general-purpose humanoid robots with human-level intelligence.

What you will do

• Own and scale the infrastructure used to train whole-body control policies, including simulation and data pipelines.
• Design fast, reliable, and highly configurable systems for controls engineers.
• Optimize hyperparameters and infrastructure to maximize training speed and model performance.
• Evaluate and integrate physics engines and simulation environments to balance realism and speed.
• Ensure high cluster utilization and minimal downtime to accelerate iteration cycles.
• Build robust tooling to move policies from training through validation to hardware deployment.

Requirements

• Production experience in Python and PyTorch.
• Experience building or scaling training infrastructure for robotics or large-scale ML workloads.
• Familiarity with physics simulation tools such as NVIDIA PhysX, MuJoCo, Warp, or PyBullet.
• Working knowledge of dynamics, controls, and robotics systems.
• Experience with reinforcement learning, imitation learning, or policy distillation.
• Experience modeling contact interactions and photorealistic simulation environments for complex manipulation.

Nice to have

• Experience with humanoid or legged robot control.
• Background in distributed systems, job schedulers, or cluster management.
• Experience deploying ML models or control policies to real-world systems.