About the job
Join us at the forefront of Amazon's sustainability initiatives to work on environmental and social advancements that support Amazon's long-term worldwide sustainability strategy. We are looking for a robotics scientist to build and operate the first autonomous materials discovery laboratory at Amazon. This role combines deep robotics expertise (motion planning, control, platform integration) with modern Physical AI approaches (vision-language-action models, sim-to-real transfer, agentic orchestration). You will design autonomous experimental workflows that integrate dexterous robotic platforms, analytical instruments, and AI-driven hypothesis generation into a closed-loop discovery pipeline — where foundation models drive hypothesis generation and experimental planning, validated on real hardware under real chemistry.
Responsibilities
- Develop, train, and benchmark robotic manipulation policies for materials synthesis and characterization using modern policy architectures (VLA architectures, diffusion policies).
- Design and execute sim-to-real transfer strategies including domain randomization, physics parameter tuning, and visual domain adaptation for laboratory robotic systems.
- Integrate robotic platforms and laboratory instruments into automated workflows via APIs (SiLA 2, or equivalent), building real-time data pipelines for multimodal experimental outputs.
- Architect policy training pipelines combining teleoperation data, synthetic demonstrations, reinforcement learning, and imitation learning for dexterous lab manipulation.
- Build production-grade agentic runtime systems — failure detection, retry logic, exception handling, and human-handoff protocols — for unattended experimental sessions.
- Design and execute autonomous experimental campaigns applying active learning, Bayesian optimization, or RL to drive iterative materials discovery.
- Drive technical design reviews and set scientific direction for the autonomous lab platform.
Qualifications
Minimum
- Master's degree, or PhD
- 3+ years of industry or academic research experience
- Knowledge of programming languages such as C/C++, Python, Java or Perl
- Experience with popular deep learning frameworks such as MxNet and Tensor Flow
Preferred
- First-hand sim-to-real transfer experience: policies trained in simulation, successfully deployed on physical hardware.
- Experience with VLA or robot policy architectures (OpenVLA, π0, RT-2, or equivalent).
- 2+ years with collaborative robot platforms including motion planning, impedance/force control, and multi-step manipulation.
- Experience building agentic AI systems for multi-step workflows including failure recovery and foundation model reasoning.
- Experience with self-driving laboratory (SDL) systems or automated chemical synthesis platforms.
- Publications in top-tier venues (NeurIPS, ICML, ICLR, ICRA, CoRL, RSS)