Deploy Embodied AI Platform CDK Stacks (N1.5)

This skill walks through deploying the two CDK stacks that make up the platform's AWS infrastructure, setting up SSH access to the GPU workstation, submitting training jobs, visualizing metrics with TensorBoard, and running closed-loop model evaluations.

For N1.6 (GR00T N1.6 / IsaacSim 5.1.0): See N16/SKILL.md. For N1.7 (GR00T N1.7 / Cosmos-Reason2-2B backbone): See N17/SKILL.md.

The two stacks are:

• IsaacGr00tBatchStack — VPC, EFS, ECR, CodeBuild, AWS Batch compute environment and job queue
• IsaacLabDcvStack — GPU-accelerated DCV workstation (EC2 instance with Elastic IP)

The DCV stack depends on the Batch stack (shares its VPC and EFS), so deploy order matters: Batch first, DCV second. Destroy order is reversed: DCV first, Batch second.

Phase 1: Prerequisites

Before deploying, verify these are in place.