Agility Robotics is looking for an experienced engineer to lead the integration of learned skill policies into their production stack, bridging upstream innovation and real-world deployment.
Requirements
- Lead the integration of AI-driven behaviors into our skill framework, ensuring they are robust, testable, and production-grade.
- Define best practices and infrastructure for evaluating, curating, and integrating learned skill policies.
- Drive the development of data collection workflows (e.g., teleop), including annotation, quality control, and feedback loops with our AI team.
- Develop and maintain high-quality interfaces between the learned skill policies and orchestration and execution layers (e.g., behavior trees, state machines, task planners).
- Build internal tools to support rapid iteration of the collect > train > evaluate > deploy loop.
- Contribute to on-robot testing and analysis to evaluate real-world performance and identify bottlenecks.
- Prior experience with imitation learning, offline RL, or learned policy deployment on humanoid robots.
Responsibilities
- Lead the integration of AI-driven behaviors into our skill framework, ensuring they are robust, testable, and production-grade.
- Define best practices and infrastructure for evaluating, curating, and integrating learned skill policies.
- Drive the development of data collection workflows (e.g., teleop), including annotation, quality control, and feedback loops with our AI team.
- Develop and maintain high-quality interfaces between the learned skill policies and orchestration and execution layers (e.g., behavior trees, state machines, task planners).
- Build internal tools to support rapid iteration of the collect > train > evaluate > deploy loop.
- Contribute to on-robot testing and analysis to evaluate real-world performance and identify bottlenecks.
- Replay and analyze failed robot trajectories to identify root causes, and collect targeted teleop data to address specific failure modes and gaps in the training dataset.
Other
- You have a BS, MS, or PhD in Computer Science, Robotics, or a related field.
- You have 5+ years of professional experience working at the interface between robotics and learning-based control, whether that's policy integration, imitation learning, or deploying RL in the real world.
- You write modern Python code and have strong software fundamentals.
- You care about system performance and reliability, and know how to build evaluation pipelines.
- You are hands-on and passionate about seeing your work running on real robots, in the real world, in production, not just in sim or in isolation.
