Reference Architecture for Azure Machine Learning
Overview
This reference architecture provides a comprehensive blueprint for building, deploying, and managing end‑to‑end machine learning solutions on Azure. It describes the recommended components, networking, security, and operational practices to ensure scalability, reliability, and compliance.
Key Components
- Azure Machine Learning Workspace – Central hub for assets, experiments, and pipelines.
- Compute Targets – Azure ML Compute, Azure Kubernetes Service (AKS), Azure Virtual Machines, and Azure Synapse Spark pools.
- Data Stores – Azure Blob Storage, Azure Data Lake Storage Gen2, Azure SQL Database, and Azure Cosmos DB.
- Model Registry & Deployment – Managed model versioning, CI/CD pipelines with Azure DevOps or GitHub Actions.
- Monitoring & Governance – Azure Monitor, Application Insights, Azure Policy, and Azure Purview.
Architecture Diagram
Sample Deployment Pipeline (YAML)
trigger:
- main
stages:
- stage: Build
jobs:
- job: Train
pool:
vmImage: 'ubuntu-latest'
steps:
- task: UsePythonVersion@0
inputs:
versionSpec: '3.9'
- script: |
pip install -r requirements.txt
python train.py --config config.yaml
displayName: 'Run training script'
- stage: Deploy
dependsOn: Build
condition: succeeded()
jobs:
- deployment: DeployModel
environment: 'ml-prod'
strategy:
runOnce:
deploy:
steps:
- task: AzureCLI@2
inputs:
azureSubscription: '$(azureSubscription)'
scriptType: 'bash'
scriptLocation: 'inlineScript'
inlineScript: |
az ml model deploy -n mymodel -f model.pkl -e myendpoint --compute aks-cluster
displayName: 'Deploy model to AKS'