CI/CD for Machine Learning
summary
Argues that ML needs the same CI/CD rigor as software and shows a concrete pipeline: Kubeflow and Azure ML plus Azure DevOps and Docker build containers, train models, register versions and deploy to Container Instances or AKS. Demo trains a tiny taco vs burrito classifier, profiles CPU/memory, fans out hyperparameters and registers artifacts in a model registry with versions. Some steps require manual gating, dataset versioning is still rough, but the pipeline runs end to end.
AI-written summary · about
chapters (28)
- 0:00Introduction
- 0:26About Sasha Rosenbaum
- 1:46Definitions
- 1:56ML
- 2:55ML vs DevOps
- 3:17Python
- 3:44DevOps
- 4:31Deep Learning
- 4:41Bad Propagation
- 5:01Data Science Work Environment
- 5:37ML vs Programming
- 6:38We climbed the Everest
- 6:52Complex
- 7:11Linear Regression
- 8:05ML Watts
- 8:23API Endpoint
- 9:05Do Models Change
- 9:31Recommendation Models
- 10:21Big Companies
- 11:46The Problem
- 12:02Today Personally Together
- 13:04Machine Learning Lifecycle
- 14:55Flow Diagram
- 17:34CubeFlow
- 18:59RML
- 19:39Demo Starts
- 30:14Questions
- 39:01Bias
related talks
57mDevOps for Machine Learning
05mCI/CD Pipelines and Machine Learning
24mFrom Model to Production: Automating the ML Lifecycle with CI/CD
34mManaging Machine Learning in Production with Kubeflow and DevOps
46mDeploying scalable Machine Learning models with Kubernetes
17mContinuous deployment for ML: The new software development lifecycle
39mSimplify, speed and improve development with DevOps
56mDevOps for Machine Learning