MLOps with Spinnaker: through the eyes of a beginner
summary
Shows how to productionize notebooks by containerize Jupyter notebooks with Docker, wire training and steps into Kubeflow pipelines for repeatable ML workflows, and run CI with Jenkins to test training and evaluation. Spinnaker handles deployments, rollbacks, and cron-triggered retrains, and model serving ties into Prometheus/Grafana for telemetry. Example workflow: code → containerized notebook → Kubeflow pipeline → Jenkins tests → Spinnaker deploy, with weekly retrain as a concrete use case.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 0:45MLOps Life Cycle
- 1:43Automating Python Notebooks
- 2:28Creating ML Pipelines
- 3:10Continuous Integration
- 3:34Deployment
- 4:32ML Pipeline
- 5:04Model Serving
- 5:45Model Monitoring
- 5:59Summary
related talks
22mMLOps at Snapchat: Continuous Machine Learning with Kubeflow & Spinnaker
17mIntroduction to Spinnaker on Kubernetes
32mMaking Spinnaker Your Own
43mAll Aboard: Transforming Software Delivery Using Spinnaker
22mBuilding a MLOps Platform with Kubeflow on GKE
45mGetting Started with Spinnaker (part 1)
48mContinuous Delivery with Spinnaker
45mMLOps and ML-Flops: going from dev to prod