DevOps for Machine Learning
summary
Shows a concrete end-to-end pipeline that actually builds, versions and deploys models: code and Jupyter notebooks in git, containerized training images, Kubeflow-triggered experiments that register models in Azure Machine Learning Service, an automated release that does profiling, smoke tests against ACI and then deploys to AKS with a pipeline triggered on model registration (manual approval optional). Covers infra-as-code for training VMs, artifact/model storage and versioning, and automated retraining on data drift plus monitoring and performance gates.
AI-written summary · about
chapters (9)
- 0:00Introduction to ML and DevOps
- 6:11Defining machine learning
- 11:56Understanding DevOps
- 16:56The ML development process
- 21:27Planning and collaboration
- 25:57Source control and infrastructure
- 32:13CI/CD for machine learning
- 37:55Testing and monitoring
- 42:55Practical implementation
related talks
56mDevOps for Machine Learning
42mCI/CD for Machine Learning
52mDevOps for AI in the Microsoft World
24mFrom Model to Production: Automating the ML Lifecycle with CI/CD
59mDevOps for Machine Learning Projects
39mSimplify, speed and improve development with DevOps
21mMastering the DevOps Machine Learning Pipeline
44mDevOps for Machine Learning - why is it different?