Principal Applied Scientist, Mobile Manipulation Robotics

Amazon Amazon · Big Tech · North Reading, MA · Applied Science

Principal Applied Scientist role focused on shipping whole-body planning, control, and optimization for mobile manipulation robots in warehouse automation. The role involves designing and deploying real-time controllers and motion planners, integrating them with existing systems, and ensuring trajectories are suitable for training foundation models. It also includes extending safety architectures.

What you'd actually do

  1. design and ship motion planners and controllers that coordinate all of Phoenix's degrees of freedom (mobile base, torso, arm, and end-effector) as a single unified system, enabling smooth, safe, and production-ready autonomous manipulation in Amazon fulfillment environments.
  2. architect whole-body motion planning pipelines (task-and-motion planning, trajectory optimization) that generate collision-free, time-efficient trajectories across the full kinematic chain.
  3. develop and tune real-time whole-body controllers, including QP-based and model-predictive approaches, and integrate them with the Foundry Controller mainline.
  4. ensuring planner-generated trajectories are high-quality enough to train downstream foundation model policies (e.g., π0.5) at scale.
  5. extend Phoenix's Control Barrier Function (CBF) safety architecture, guaranteeing constraint satisfaction from planning through execution.

Skills

Required

  • PhD in Robotics, Computer Science, Mechanical Engineering, or a related field, with 7+ years of relevant research experience after degree; or Master's degree with 12+ years of equivalent experience
  • Experience developing and deploying real-time controllers on physical robotic hardware
  • Demonstrated ability to influence technical strategy across multiple teams and organizations

Nice to have

  • Proficiency in Python and C++ with experience writing production-grade code
  • Proven track record of publications at top-tier venues (e.g., CVPR, ICCV, ECCV, NeurIPS, ICRA, RSS, CoRL)
  • Experience with real-time perception on resource-constrained hardware (edge compute, embedded GPUs)
  • Track record of building perception systems that generalize across multiple sensor configurations or robot platforms
  • Experience with foundation models or large-scale self-supervised learning applied to robotics perception

What the JD emphasized

  • production-ready autonomous manipulation
  • real-time whole-body controllers
  • train downstream foundation model policies

Other signals

  • shipping production-ready autonomous manipulation
  • whole-body planning, control, and optimization stack
  • training downstream foundation model policies