Lightning Talk: Kuber-What-Es?! Misadventures in Building UIs for K8s-Based ML Platforms
Lightningat KubeCon + CloudNativeCon North America 2019 · on Tuesday, 2019-11-19· at 17:20· Exhibit Hall AB - San Diego Convention Center, Ground Level·
summary
Team builds Kubernetes-based ML platform to run massively parallel hyperparameter optimization jobs and chooses three UI rules: focus on non-infra users, CLI with hidden Kubernetes. After launch usage splits: data scientists want notebook integrations and no terminals while ML engineers want direct CLI and tinkering, so the product pleases neither and adoption is low. Conclusion is to observe actual workflows, pick one persona at a time, and iterate on concrete integrations rather than one-size-fits-all abstractions.
AI-written summary · about
chapters (5)
related talks
12mLightning Talk: Benchmarking Your Distributed ML Training on the K8s Platform
42mThe Challenges Managing a Kubernetes-Based Machine Learning Infrastructure
33mFrom Chaos To Control: Building ML Platform
22mBuilding a MLOps Platform with Kubeflow on GKE
12mKeynote: Building Enterprise AI/ML Platform on Kubernetes
30mBuilding ML Products With Kubeflow
10mChallenges and Opportunities in Making AI Easy and Efficient with...
37mTowards Kubeflow 1.0, Bringing a Cloud Native Platform For ML to Kubernetes