Applied Scientist, Safe RL for Dynamic Legged Robots

Amazon; San Francisco (CA); On-site; USD 143,000 - 193,000; Full time

Benefits offered by this job

Health insurance

401(k) matching

Paid time off

Job summary

Amazon's SAF Lab seeks an Applied Scientist to advance safe reinforcement learning for legged locomotion on physical robots. You will develop RL architectures interfacing with physics models, internalize safety constraints in training, and enable sim-to-real transfer with safety filters during run-time.

You will collaborate across perception, planning, and controls teams to push the boundaries of safe autonomy and robust robot operation around humans.

Qualifications

• PhD in Computer Science, Robotics, Mechanical or Electrical Engineering with RL focus.
• Experience applying RL to physical robotic systems, including sim-to-real transfer.

Responsibilities

• Collaborate with product teams and science leaders to set a science roadmap for real robots.
• Design, train, and deploy RL policies for dynamic legged locomotion on physical robots.
• Develop sim-to-real transfer pipelines robust to the reality gap, using domain randomization and system identification.
• Integrate control-based methods with RL, including safety constraints in training and safety layers in deployment.
• Develop and maintain large-scale training infrastructure for locomotion policy learning.

Skills

Python

PyTorch

JAX

Reinforcement Learning

Robotics

Dynamics

Education

PhD in CS/Robotics/EE

Tools

Isaac Gym/Sim

MuJoCo

PyBullet

Job description

Amazon's SAF Lab seeks an Applied Scientist to advance safe reinforcement learning for legged locomotion on physical robots. You will develop RL architectures interfacing with physics models, internalize safety constraints in training, and enable sim-to-real transfer with safety filters during run-time.

You will collaborate across perception, planning, and controls teams to push the boundaries of safe autonomy and robust robot operation around humans.