- Define a focused Physical AI research and engineering roadmap with clear hypotheses and milestones.
- Select, train, adapt, or evaluate model families for embodied tasks.
- Personally design and train Transformer-based systems for robot policies.
- Establish reproducible offline and real-world evaluation systems.
- Protect evaluations against leakage, overfitting, and misleading criteria.
- Analyze failures across data, perception, models, control, and hardware.
- Design safe, staged physical testing and deployment plans.
- Recruit and lead a small team of researchers and engineers.
- Translate customer and partner needs into testable technical requirements.
PythonMachine LearningPyTorch+1 more