Applied Scientist – Large-Scale RL for Robotic Manipulation

Amazon; Seattle, Northern (WA, KY); Hybrid; USD 143,000 - 193,000; Full time

Benefits offered by this job

Health insurance

401(k) matching

Paid time off

Parental leave

Job summary

Amazon Robotics seeks an Applied Scientist to advance reinforcement learning for manipulation. You will design policies for non-prehensile, contact-rich tasks, build large-scale simulation environments, and transfer policies to physical robots.

Collaboration with control, perception, and hardware teams will be essential as you publish and present results in academia. The role emphasizes creating robust, scalable RL workloads, with production-quality code and a focus on sim-to-real transfer.

Qualifications

• PhD or MS with 4+ years in science, tech, engineering or related field.
• Experience programming in Java, C++, Python or related language.
• Experience with patents or publications at top-tier conferences or journals.
• Experience training RL or imitation learning policies for manipulation or robot control.
• Strong background in RL, including reward design and training at scale in simulation.

Responsibilities

• Design and train RL policies for non-prehensile and contact-rich manipulation.
• Build and scale simulation environments, curricula, and reward formulations.
• Develop sim-to-real methods and validate on physical hardware.
• Write production-quality code and own scalable training/evaluation pipelines.
• Evaluate policy behavior and iterate between simulation and real-world testing.
• Collaborate with control, perception, and hardware teams to move ideas to demonstration.
• Represent Amazon Robotics in academia through publications and presentations.

Skills

PhD or MS

C++

Python

Reinforcement learning

Publications

Sim-to-real

Education

PhD or MS in ML/CS/Robotics

Tools

Domain randomization

System identification

Job description

Amazon Robotics seeks an Applied Scientist to advance reinforcement learning for manipulation. You will design policies for non-prehensile, contact-rich tasks, build large-scale simulation environments, and transfer policies to physical robots.

Collaboration with control, perception, and hardware teams will be essential as you publish and present results in academia. The role emphasizes creating robust, scalable RL workloads, with production-quality code and a focus on sim-to-real transfer.