From Model to Production: Automating the ML Lifecycle with CI/CD
summary
Shows a runnable CI/CD MLOps pipeline that pulls code from GitLab CI, lints Python, versions data with DVC, tracks experiments with MLflow or replaces both using Azure Machine Learning workspace, builds Docker images to ACR and deploys to AKS via Helm. Includes a flight meal-demand case study that trains, evaluates and monitors models with App Insights, claims ~1% waste reduction (translating to millions saved) and surfaces drift detection, cost and retraining as the main operational challenges.
AI-written summary · about
chapters (10)
- 0:00Introduction to ML lifecycle
- 4:54Why automate ML workflows
- 5:55CI/CD pipeline for ML
- 10:13Docker and container registry
- 12:23MLOps pipeline example
- 13:48Azure ML workspace benefits
- 16:17Flight demand forecasting case
- 17:39Key challenges in MLOps
- 20:06Best practices for MLOps
- 20:48Q&A session
related talks
23mAccelerating Model Deployment with MLOps
57mDevOps for Machine Learning
45mMLOps and ML-Flops: going from dev to prod
31m🤑 Paying Off Technical Debt in the Development of Machine Learning Apps
42mCI/CD for Machine Learning
25mBuilding Scalable MLOps Pipelines That Actually Work
56mDevOps for Machine Learning
39mContinuous Delivery for Machine Learning — The Future of MLOps