about 4 hours ago
Responsibilities
- Implement and deploy state-of-the-art RL algorithms for dynamic locomotion and manipulation tasks.
- Drive the development cycle from simulation prototyping to policy fine-tuning on robots.
- Optimize and scale the RL training pipeline for faster iteration and high-throughput simulation.
- Develop motion retargeting pipelines to convert human demonstration data into reference trajectories.
- Collaborate with robotics and hardware teams to diagnose system-level issues and co-develop solutions.
- Analyze and present hardware results to guide technical directions and demonstrate progress.
Requirements
- 3+ years of hands-on expertise with RL frameworks like PyTorch or JAX.
- Proficiency in high-fidelity physics simulators such as MuJoCo or IsaacGym.
- Strong mastery of Python for rapid prototyping and C++ for performant code development.
- Experience with large-scale, distributed training pipelines and their optimization.
- Deep theoretical understanding of modern reinforcement learning and related techniques.
- Strong intuition for robot dynamics and controls theory.