Develop a Robot Policy with Agent Skills and NVIDIA Cosmos 3 Edge on Viam | Cosmos Labs 

Learn how to build and validate a robot arm policy without a task-specific real-world dataset. In this livestream, we’ll speed-run creating a simulation of your robot and task, then training a motion policy in sim, and finally deploying the policy back to your sim for testing. You’ll leave with a clear understanding of the end-to-end workflow for creating, training, and testing a Cosmos 3 motion policy and be ready to try one yourself. You’ll also learn how to use Cosmos and Viam Robotics agentic infrastructure to train, customize, and deploy robots safely.  
What You'll Learn:

• How to create a synthetic dataset for training 
• How to use the spatial reasoning capacity of frontier models to design your training tasks
• How to train and run Cosmos 3 policies

Have questions about Cosmos 3 policies, Isaac sim, designing with agents, or Viam? Drop them in live — the NVIDIA and Viam teams will address them during the discussion.

📄 Read the technical blog →https://nvda.ws/4ibLaF3 
📚 Explore the NVIDIA Cosmos platform → https://github.com/nvidia/Cosmos
👥 Join the NVIDIA Omniverse community → https://discord.com/invite/nvidiaomniverse