Applied Scientist — Real-Robot Manipulation & Learning

Amazon Science; Seattle (WA); On-site; USD 143,000 - 193,000; Full time

Job summary

Amazon Science in Seattle seeks an Applied Scientist to bring learned manipulation policies to life on real robots. You will focus on making learned, contact-rich, non-prehensile behaviors work reliably on hardware, closing the loop between policy and real contact, sensing, and dynamics.

You will join a small team advancing manipulation capabilities across products, with opportunities to publish and collaborate with control, perception, and hardware experts.

Qualifications

• 3+ years of building models for business application experience.
• PhD, or Master's degree and 4+ years of science, technology, engineering or related field experience.
• Experience in patents or publications at top-tier peer-reviewed conferences or journals.
• Experience programming in Java, C++, Python or related language.
• Hands-on experience deploying and evaluating learning-based control or manipulation on physical robots.
• Strong background in one or more of: reinforcement learning, learning-based manipulation, contact-rich control, or force/torque control.

Responsibilities

• Deploy and evaluate learned manipulation policies on physical robots, focused on the long tail of diverse, demanding conditions.
• Diagnose the gap between simulated and real behavior, and drive the changes in policies, rewards, sensing, or control that close it.
• Develop tactile, actuation, and contact-aware methods that make non-prehensile manipulation fast, assured, and robust.
• Write production-quality code and own scalable, real-time implementations that run on robots.
• Build the data collection, logging, and evaluation loops that turn real-world experience into policy improvements.
• Partner with scientists and engineers across control, perception, and hardware to move ideas from prototype to demonstrated capability on hardware.

Skills

Java

C++

Python

Reinforcement learning

Learning-based manipulation

Contact-rich control

Force/torque control

Patents or publications

Education

PhD

Masters + 4+ years experience

Job description

Amazon Science in Seattle seeks an Applied Scientist to bring learned manipulation policies to life on real robots. You will focus on making learned, contact-rich, non-prehensile behaviors work reliably on hardware, closing the loop between policy and real contact, sensing, and dynamics.

You will join a small team advancing manipulation capabilities across products, with opportunities to publish and collaborate with control, perception, and hardware experts.