3 days ago
Responsibilities
- Design, build, and maintain end-to-end motion data pipelines integrating sources like mocap and teleoperation.
- Implement and optimize kinematic and dynamic retargeting pipelines for accurate human demonstration mapping.
- Develop tools and scripts for processing and cleaning raw human demonstration data.
- Leverage game engines and physics simulators for procedural motion generation and data augmentation.
- Generate high-volume, high-quality trajectory datasets for training reinforcement learning policies.
- Write robust, automated pipelines to streamline data flow between various sources and training infrastructure.
- Collaborate with Reinforcement Learning and Controls teams to ensure generated trajectories are physically viable.
Requirements
- Strong understanding of robot kinematics, coordinate transformations, and rigid body dynamics.
- Experience building or maintaining pipelines for spatial data, including motion capture and teleoperation.
- Expertise in Python for scripting, data processing, and pipeline automation.
- Results-oriented mindset with eagerness to bridge human motion data and robotic control algorithms.
- High adaptability and willingness to explore new tools across the robotics software stack.