E2E ML Platform on Kubernetes with a few clicks
summary
Demonstrates getting a full Kubeflow E2E platform up on Kubernetes using the Kubeflow Operator to automate deploy, monitoring and lifecycle management, installing the 1.3 manifests and operator in minutes. Shows Jupyter/VS Code/RStudio notebooks, Argo or Tekton pipelines, KFServing for model serving, Katib for HPO, Istio+Prometheus+Grafana for monitoring and MLMD for metadata. Notes multi-tenant namespaces, portability from local to cloud, and warns about component upgrade complexity.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:56ML evolution and MLOps rise
- 4:29End-to-end ML platform
- 8:21Cloud ML platforms pros and cons
- 11:53DIY vs open source ML platforms
- 14:24Desired ML platform features
- 15:48Kubeflow as the solution
- 17:54Kubeflow scalability and composability
- 19:39Portability and consistency in Kubeflow
- 21:44Kubeflow components overview
- 23:52Kubeflow 1.3 improvements
- 24:54Kubeflow operator challenges
- 26:58Kubeflow future and features
- 27:40Live Kubeflow demo
- 30:04Q&A
related talks
32mE2E ML Platform on Kubernetes with just a few clicks
01mE2E ML Platform on Kubernetes with just a few clicks
39mDeploy your ML Workloads on Kubernetes with Kubeflow
14mKubeflow: Where ML Meets Kubernetes
22mBuilding a MLOps Platform with Kubeflow on GKE
46mDeploying scalable Machine Learning models with Kubernetes
51mMachine Learning using Kubeflow and Kubernetes
28mMachine Learning on Kubernetes at Shell: A Kubeflow Journey