onsite
Robot Learning Engineer - 9series
Software Engineer
Lead end‑to‑end robot learning pipelines, crafting synthetic datasets in Isaac Lab/Mimic, integrating Cosmos for visual realism, and training advanced vision encoders and diffusion/flow‑matching action heads to deploy robust policies.
About the role
Key Responsibilities
- Design, implement, and maintain synthetic data pipelines in Isaac Lab and Isaac Mimic, including scene and asset setup, demonstration generation, and domain randomization.
- Integrate Cosmos or similar world‑foundation‑model tooling to enhance visual realism and reduce the sim‑to‑real gap.
- Build and tune co‑training pipelines that blend real demonstrations with synthetic data for optimal policy learning.
- Develop and fine‑tune custom vision encoders, diffusion and flow‑matching action heads, and VLA finetunes to drive high‑performance robot policies.
- Collaborate with research and engineering teams to iterate on data generation strategies and model architectures.
Requirements
- Strong programming skills in Python and experience with robotics simulation frameworks.
- Hands‑on experience with Isaac Lab, Isaac Mimic, or equivalent simulation environments.
- Proficiency in synthetic data generation, domain randomization, and visual realism techniques.
- Knowledge of vision encoder architectures, diffusion models, and flow‑matching action heads.
- Excellent problem‑solving skills and ability to work cross‑functionally in a fast‑paced environment.