Senior Robotics Scientist: Dexterous Grasping & Policies

Amazon Science; Sunnyvale (CA); On-site; USD 192,000 - 260,000; Full time

Benefits offered by this job

Health insurance

RSUs

Paid time off

Parental leave

Job summary

Amazon Science seeks top talent to advance tactile- and force-driven manipulation for multi-fingered robotic hands. You will research policy learning methods and deploy them on real hardware, bridging simulation and physical systems with a focus on safe, dexterous manipulation in unstructured environments.

The role emphasizes hands-on research, collaboration across hardware and controls teams, and publishing results in top venues while advancing toward real-world robotic deployment.

Qualifications

• PhD or Master's degree and 6+ years of applied research experience.
• Experience programming in Java, C++, Python or related language.
• Research or applied experience in tactile sensing, manipulation policy learning, or multi-modal perception for robotics.

Responsibilities

• Design and train manipulation policies that leverage tactile and force feedback for dexterous grasping and in-hand manipulation.
• Drive sim-to-real transfer between simulation and real robotic hands.
• Shape objectives and control strategies with hardware constraints in mind.
• Build simulation-based and on-robot evaluation frameworks with clear benchmarks.

Skills

Java

C++

Python

Applied robotics research

Reinforcement learning

Education

PhD or MSc with 6+ years of experience

Tools

Robot hardware deployment

MuJoCo/Isaac

URDF

TensorFlow/PyTorch

Job description

Amazon Science seeks top talent to advance tactile- and force-driven manipulation for multi-fingered robotic hands. You will research policy learning methods and deploy them on real hardware, bridging simulation and physical systems with a focus on safe, dexterous manipulation in unstructured environments.

The role emphasizes hands-on research, collaboration across hardware and controls teams, and publishing results in top venues while advancing toward real-world robotic deployment.