Managing Machine Learning in Production with Kubeflow and DevOps
summary
Live demo builds three container images via Azure Pipelines on every merge, triggers a Kubeflow pipeline that runs MobileNet transfer learning on GPUs, registers artifacts to an Azure model registry service, and promotes a canary then production endpoint after human validation. The CI/CD run count reaches the eighties during debugging and completes successfully, showing reproducible artifacts, metadata, and audit trails that enable continuous training and portable Kubeflow pipelines across clouds.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 3:24ML is hard
- 6:48ML Ops
- 9:45Building your own ML Ops
- 10:37Multicloud ML Ops
- 13:41Demo
- 21:09Service Endpoint
- 26:29Questions
related talks
42mCI/CD for Machine Learning
57mDevOps for Machine Learning
46mDeploying scalable Machine Learning models with Kubernetes
46mAccelerating Machine Learning DevOps with Kubeflow
36mMachine learning pipeline Kubeflow, let's start
05mCI/CD Pipelines and Machine Learning
39mOperating Deep Learning Pipelines Anywhere Using Kubeflow
54mScaling Machine Learning Pipelines in Cloud