E2E ML Platform on Kubernetes with just a few clicks
summary
Claims an end-to-end ML platform on Kubernetes can be provisioned with a few clicks by installing Kubeflow via the manifests or the Kubeflow Operator, then composing Jupyter/code servers, Argo/Tekton pipelines, training runtimes (TensorFlow/PyTorch/XGBoost), KFServing for model serving, and Istio + Prometheus/Grafana for monitoring. Demo deployment runs on IBM Cloud with per-user namespaces, a sample pipeline (coin‑flip) that trains and serves, and portability from Kubeflow 1.3 local minikube to cloud clusters.
AI-written summary · about
chapters (12)
- 0:00Start
- 5:47End-to-End Machine Learning Platform
- 6:23Data Ingestion
- 9:18Pros of Using Enter Machine Learning Platform
- 10:16Cons
- 22:19Keyflow Components
- 23:56Manifest Repository
- 25:30Keyflow Operator
- 27:42Future
- 28:15References
- 28:39Key Flow Environment
- 29:01Kickflow Dashboard
related talks
30mE2E ML Platform on Kubernetes with a few clicks
39mDeploy your ML Workloads on Kubernetes with Kubeflow
14mKubeflow: Where ML Meets Kubernetes
15mKeynote: Machine Learning on Kubernetes Made Easy With Kubeflow
46mDeploying scalable Machine Learning models with Kubernetes
30mKubeflow, Your ML Tools on Kubernetes
51mMachine Learning using Kubeflow and Kubernetes
22mBuilding a MLOps Platform with Kubeflow on GKE