MLOps and ML-Flops: going from dev to prod
summary
Shows how to move from messy local notebooks (hundreds of Untitled_final.ipynb files) to a repeatable dev/UAT/production pipeline using Azure services: Databricks + Delta lake medallion, Azure Machine Learning for model/version tracking, Key Vault and App Insights for secrets and monitoring, and Terraform to codify ~100+ setup steps. Demonstrates CI/CD with pipelines that trigger automated retrain when data or model drift appears, and containerized endpoints for production deployment.
AI-written summary · about
chapters (7)
- 0:00Start
- 3:441. MLOps 2. MLflops 3. Architecture
- 5:49What's the goal? We want version-able, deployable, and responsible machine learning projects that we can automatically retrain and monitor
- 9:06automatic training
- 15:08Need a DevOps solution that can handle the unique challenges associated with Machine Learning
- 18:13Repeatability Set-up times
- 24:22Machine Learning Data and model drift Data and model versioning Model Retraining
related talks
24mFrom Model to Production: Automating the ML Lifecycle with CI/CD
27mBeginners guide to Azure MLOps
57mDevOps for Machine Learning
51mHow to get started with MLOps in Azure Databricks
13mMLOps: From AI Research to Production-Ready Systems
47mMLOps tools to automate ML deployment
48mStreamlining your Machine Learning projects with Azure Machine Learning
32mProductionize Machine Learning on Azure