LeRobot, Policy Training & Real-World Deployment
Collect datasets in simulation and on real robots, train manipulation policies with LeRobot, and deploy learned models back onto physical arms.
Course overview
The course culminates in data-driven robot learning. Students collect LeRobot-format datasets from teleoperation and simulation, train manipulation policies, study common model families, and deploy learned models back to the physical arm.
Core curriculum
Four themed modules. Each module is a working block of lessons and labs.
Dataset Creation
Use Isaac Sim and real-robot demonstrations to build structured datasets with observations, actions, rewards, and task metadata.
Policy Training
Train and evaluate LeRobot policies for manipulation, including imitation-learning baselines and modern policy architectures.
Model Families
Compare common model families for robotics, such as ACT, diffusion-style policies, and lightweight vision-language-action approaches.
Deployment & Eval
Deploy trained models onto the arm with safety wrappers, runtime monitoring, and real-world evaluation of success, latency, and robustness.
What you'll gain
- LeRobot-format datasets from sim and real robots
- Trained manipulation policies you can demo
- Hands-on comparison of modern robotics model families
- A real-world deployment + evaluation report
Available classes
Open classes you can enroll in directly. Each class shows its instructor and weekly schedule.
Not sure which class fits? Reach us from the Contact page.
Ready to build with AI?
Schedule a free consultation. We'll help you choose the right track — Builder, Portfolio, Scholar, or Innovation — based on your goals and experience.