How microsoft/ai works
Project type and purpose
• This is not a single application but a “meta-repository” that organizes Microsoft’s open-source AI content:
• AI100 (samples)
• AI200 (reference architectures)
• AI300 (best practices)
• It also contains CI/CD pipelines (.ci folder), Azure ARM templates (AzureDeployment folder), documentation (.docs/.images), and supporting utilities.
File structure
• Top-level components:
• .ci – Azure DevOps pipeline definitions, scripts, variables, stages & steps
• .github – issue templates and GitHub metadata
• .docs & .images – markdown guides and assets
• AzureDeployment – ARM templates for CosmosDB, ML services, Synapse, Identity
• ai100-samples, ai200-architectures, ai300-practices – submodules of sample/reference/practice repos
• submodules – a mirror of the above for Git submodule management
• utilities – shared scripts or code used by samples & pipelines
• Heavy use of Azure DevOps YAML for CI, and ARM for IaC.
README for additional insights
• Purpose: central hub of AI samples (AI100), reference architectures (AI200), best practices (AI300)
• Workflow: users clone with submodules, pick scenarios/samples, follow docs for deployment/training/scoring.
Elements
• Clear labels: “GitHub Repo,” “CI/CD Engine,” “ARM Templates,” “Submodule: AI100 Samples,” etc.
• From GitHub to CI/CD trigger
• From CI/CD to Azure (resource provisioning)
• From CI/CD to PyPI/Docs site
• Rectangles for code components (submodules, utilities)
• Hexagons for cloud services (AML, CosmosDB, Synapse)