Y Combinator Research Scientist
Summary
Research Scientist at Ultra builds and fine-tunes robot foundation models to improve task performance and reliability for industrial deployments, integrating research with production robotics stacks.
About the role
We are looking for high-impact researchers and engineers to help bridge the gap between robot foundation models and production robot deployment. Speed and reliability are the two metrics that matter the most for industrial customers, yet most robot models struggle on both. You will post-train robot foundation models and do new research to hill climb task performance in our research lab and at real customer deployments.
What You’ll Do
- Improve our robots’ ability to quickly learn new tasks and to execute those tasks quickly and reliably
- Lead research efforts around improving our models’ robustness and performance for real world tasks
- Integrate trained models with our production robotics stack and work with deployment teams to scale models across real customer sites
Who You Are
- Have experience with fine-tuning large models for downstream tasks (IL, RL, etc.)
- Operate with high agency and are able to own significant research efforts end to end
- Care deeply about evaluation, benchmarking, and failure analysis
- A generalist who enjoys getting your hands dirty across the whole stack ML stack
- Passionate about seeing robots make an impact in the real world
Bonus Points
- Experience applying ML to robotics, with hands-on robot experience
- MS/PhD in Computer Science, Machine Learning, Robotics, or a related field