About This Architecture

Enterprise MLOps pipeline deploys trained models from Azure Databricks MLflow to OpenShift on Azure via AKS. Azure Databricks trains models tracked in MLflow Tracking Server, registers artifacts in ML Model Registry, and stores binaries in Azure Storage Account. Azure DevOps CI/CD Pipeline pulls containerized models from Azure Container Registry, deploys to AKS Cluster with VM Scale Sets Worker Nodes running Container Instances ML Pods. App Service API Endpoint serves predictions secured by Key Vault Secrets, fronted by Azure Load Balancer and API Management for client access. Azure Monitor, Application Insights, and Log Analytics provide observability across the deployment pipeline. Fork this diagram on Diagrams.so to customize your Azure MLOps architecture, export as .drawio or .png, or embed in runbooks.

People also ask

How do I deploy MLflow models from Azure Databricks to OpenShift on AKS with CI/CD?

This architecture shows MLflow models trained in Azure Databricks, registered in ML Model Registry, containerized in Azure Container Registry, and deployed to OpenShift AKS via Azure DevOps CI/CD Pipeline. Container Instances serve predictions through App Service API Endpoint fronted by API Management, with Key Vault for secrets and Azure Monitor for observability.