Building a Medical AI with Kubernetes and Kubeflow
Talkat KubeCon + CloudNativeCon North America 2019 · on Thursday, 2019-11-21· at 15:20· Room 11AB - San Diego Convention Center, Upper Level·
summary
Moves heavy ML workloads from bare metal into Kubernetes and Kubeflow, combining Kubernetes, Kubeflow, Argo, Istio, OPA, Flux, Prometheus, Grafana, Vault to deliver an auditable, multi-region research platform; projects map to isolated namespaces/profiles with shared volumes, per-project dashboards and cost metrics (kube-cost → Prometheus). Runs ~50 cloud nodes for large hyperparameter searches (multiple TB RAM) and parallelizes clinical validation from about 10 hours down to ~20 minutes. Adds an API gateway for multi-cluster orchestration, GPU controls, and plans KFServing/KNative and MLflow integration.
AI-written summary · about
chapters (21)
- 0:00Intro
- 0:40Welcome
- 1:07Introduction
- 2:35The affordability problem
- 3:50How this all started
- 5:05Deploying a cluster
- 5:44Why Kubernetes
- 7:25Overview
- 7:56Infrastructure
- 10:58Networking
- 12:14Cluster management
- 13:44Kubeflow
- 15:50Accessing data
- 17:10Monitoring
- 21:21API
- 22:27Complex objects
- 23:37Demo
- 24:49Use cases
- 25:40Clinical validation
- 28:01Next steps
- 29:37Wrap up
related talks
46mDeploying scalable Machine Learning models with Kubernetes
51mMachine Learning using Kubeflow and Kubernetes
54mEconomics and Best Practices of Running AI/ML Workloads on Kubernetes
31mKubernetes for Machine Learning: Managing and Scaling AI Workloads
14mKubeflow: Where ML Meets Kubernetes
31mBuilding and Managing a Centralized ML Platform with Kubeflow at...
39mDeploy your ML Workloads on Kubernetes with Kubeflow
27mKubeflow for enabling AI Powered Drug Discovery and development...