Workshops & Resources¶
owner: Youngjin · updated: 2026-09 · volatility: medium
L0 TL;DR: Public self-paced workshops, implementation guides, and samples in one place. Last checked: 2026-09-19. Open each title for the source and use the related pillar for architecture context.
Duration and cost depend on modules, account readiness, and GPU availability. Check source Prerequisites/Cleanup, permissions, and model terms before starting. This list records source inspection, not a full rerun of every lab.
Choose by goal¶
| Goal | Resource | Related pillar |
|---|---|---|
| Start simulation training | NVIDIA Isaac Lab on AWS | P3 |
| RL to VLA1 in Korean | Physical AI E2E workshop | P2 |
| Agent component fundamentals | Getting Started with Amazon Bedrock AgentCore | P5 |
| Advanced AgentCore application | Diving Deep with Amazon Bedrock AgentCore | P5 |
| Read π0 training/evaluation setup | π0 fine-tuning on SageMaker HyperPod EKS | P2 |
| Connect 3D assets, sensors, and prediction | OpenUSD industrial digital-twin sample | P3 |
| Inspect simulation results without a robot | VLA Simulator on AWS | P4 |
| Collect physical robot demonstrations | LeRobot data collection on Greengrass | P1 |
NVIDIA Isaac Lab on AWS¶
NVIDIA Isaac Lab on AWS · Official AWS workshop · English
Content: Run Isaac Lab on EC2, scale with AWS Batch, inspect the model in Isaac Sim, then clean up.
Before starting: Prepare AWS access, container basics, and GPU quota. The exercise is simulation-based.
Related pillar: P3.
Physical AI E2E workshop¶
Physical AI E2E workshop · Community-authored guide · Korean
Content: Modular exercises for Isaac Lab RL, GR00T VLA, Batch/SageMaker, and simulation evaluation.
Before starting: Check CDK, Docker, GPU setup, and model access. Choose the needed track first.
Related pillar: P2.
Getting Started with Amazon Bedrock AgentCore¶
Getting Started with Amazon Bedrock AgentCore · Official AWS workshop · English
Content: Progress through an agent prototype, Memory, Gateway, observability/evaluation, and Policy.
Before starting: Check AWS account and model/service access. Robot connectivity needs separate integration.
Related pillar: P5.
Diving Deep with Amazon Bedrock AgentCore¶
Diving Deep with Amazon Bedrock AgentCore · Official AWS workshop · English
Content: Explore composition, security, and observability through Feature Deep Dive or Build with Skills.
Before starting: The workshop requires Getting Started first; follow its development-environment prerequisites.
Related pillar: P5.
π0 fine-tuning on SageMaker HyperPod EKS¶
π0 fine-tuning on SageMaker HyperPod EKS · AWS implementation article · English
Content: New guide connecting data, training, evaluation, and cleanup on HyperPod EKS; published 2026-09-10.
Before starting: Advanced setup. The article is public, but its code link (awsome-distributed-ai/…/3.test_cases/pytorch/pi0-lerobot) returns 404 — the code was never merged to main and sits in unmerged PR #1174 (checked 2026-10-14; PR last updated 2026-07-30). Unmerged code is unreviewed, so use this as reading material.
Related pillar: P2.
OpenUSD industrial digital-twin sample¶
OpenUSD industrial digital-twin sample · aws-samples · English
Content: L1 static scene, L2 sensor overlays, L3 prediction, and L4 recalibration; local and AWS execution paths.
Before starting: Follow Python/Node.js/Docker prerequisites. Clone from the public GitHub URL linked here instead of the internal address in the README example.
Related pillar: P3.
VLA Simulator on AWS¶
VLA Simulator on AWS · aws-samples · English
Content: Runs supported VLA/simulator combinations and collects videos and summaries in S3.
Before starting: Check CDK, GPU quota, and per-model licenses. Small demo success rates do not establish customer deployment performance.
Related pillar: P4.
LeRobot data collection on Greengrass¶
LeRobot data collection on Greengrass · aws-samples · English/Korean
Content: Records SO-ARM101 leader/follower and camera data in LeRobot format and uploads it to S3.
Before starting: Requires robot hardware, cameras, and a Greengrass device. The README marks it educational/demo code, not ready for production.
Related pillar: P1.
Suggested learning order¶
For simulation, start with Isaac Lab → the relevant E2E track → VLA Simulator. For agents, follow AgentCore Getting Started → Deep Dive. For digital twins, start with local OpenUSD execution → the AWS levels you need.
Maintaining this list¶
Record provider, language, goal, prerequisites, related pillar, and check date for new resources. Link availability and lab reproducibility are separate checks. Use News for announcements and Radar for technology awaiting validation.
-
VLA (Vision-Language-Action) — A model that turns visual observations and language instructions into robot actions. ↩