Documentation
Technical documentation for deploying, training, and operating robotics workloads on Azure with NVIDIA Isaac and OSMO. This index organizes every guide, reference, and walkthrough in the repository by topic so you can find what you need based on where you are in the workflow.
Documentation spans the full lifecycle, from provisioning Azure infrastructure with Terraform, through training reinforcement-learning policies with Isaac Lab and AzureML, to running inference on edge devices. Each section targets a specific audience and phase of the project.
👤 Audience Guide
| Role | Start here |
|---|---|
| First-time deployer | Getting Started, then Deployment Guide |
| ML / Robotics engineer | Training and Inference (coming soon) |
| Platform operator | Operations and Security Guide |
| Contributor | Contributing |
🪜 Tier Guide
Adoption is modeled as six graduated tiers (T0-T5), each a legitimate stopping point. T0 — Dev is the default starting path (one laptop, one robot, zero cloud, zero Kubernetes). T2 — Pilot is the recommended production path. T3-T5 are advanced and opt-in. Pick the tier that matches your reach, then follow its quick-start and read its infrastructure boundaries. See the canonical Tier Model for the authoritative tier table and vocabulary.
| Tier | Scope | Quick start | Architecture |
|---|---|---|---|
| T0 — Dev ⭐ | Laptop + 1 robot, zero cloud, zero Kubernetes | Tier 0 — Dev | T0 — Dev |
| T1 — Lab | One site, a few robots, shared GPU; first cloud storage | Tier 1 — Lab | T1 — Lab |
| T2 — Pilot ✅ | One site at scale; cloud training default | Tier 2 — Pilot | T2 — Pilot |
| T3 — Production | Single-site declarative deploy (local k3s + Flux, no Arc) | Tier 3 — Production | T3 — Production |
| T4 — Scale | Multi-site fleet delivery; Arc reachability broker | Tier 4 — Scale | T4 — Scale |
| T5 — Operate | Fleet intelligence for drift detection and retraining | Tier 5 — Operate | T5 — Operate |
⭐ default · ✅ recommended production
[!NOTE] Roadmap honesty. T5 (Operate / fleet intelligence) is on the roadmap and not yet available. The fleet-intelligence domain is currently specified, with implementation planned. Today's shipping capability spans T0-T4.
📖 Documentation Index
| Section | Description | Status |
|---|---|---|
| Getting Started | Environment setup, prerequisites, and first deployment walkthrough | Available |
| Deployment Guide | Infrastructure provisioning with Terraform, AKS cluster setup, and networking | Available |
| Training | Model training pipelines with Isaac Lab, AzureML jobs, and OSMO orchestration | Available |
| Inference | Serving trained policies for real-time control on edge and cloud | Coming soon |
| Workflows | AzureML and OSMO job templates, pipeline configuration, and submission scripts | Coming soon |
| Operations | Monitoring, scaling, troubleshooting, and cost management for running clusters | Available |
| Security | Identity, networking, compliance, and hardening for production deployments | Available |
| Reference | CLI parameter tables, script usage, workflow templates, and configuration reference | Available |
| Contributing | Contribution guidelines, PR process, deployment validation, and coding conventions | Available |
📄 Current Guides
Standalone guides available now. These cover common tasks and will move into their respective topic sections as the documentation structure expands.
| Guide | Description |
|---|---|
| MLflow Integration | Configuring MLflow experiment tracking for SKRL training agents with automatic metric logging to Azure ML |
| Security Guide | Security configuration inventory, deployment responsibilities, and hardening checklist for robotics workloads |
🚀 Next Steps
- Review the deployment guide for infrastructure provisioning and cluster setup
- Explore MLflow Integration to set up experiment tracking for training runs
- Read the Contributing guide to get involved with the project
🤖 Crafted with precision by ✨Copilot following brilliant human instruction, then carefully refined by our team of discerning human reviewers.