Panel: Enterprise-grade, On-prem Kubeflow in the Financial Sector
Panelat KubeCon + CloudNativeCon North America 2019 · on Wednesday, 2019-11-20· at 14:25· Room 14AB - San Diego Convention Center, Mezzanine Level·
summary
Enterprise Kubeflow on-prem details: deployments use Arrikto-style local-disk provisioning with dynamic local PV snapshots exported to object storage for portability, cutting a TF pipeline from 26→12 steps in six months and eliminating Kubernetes expertise for data scientists; pipeline UI and notebooks dominate usage, Feast and KFServing are upcoming; teams report 40% less non-ML code due to pipeline automation but hit real blockers: air-gapped install hurdles (internal Docker repo layout, proxy, admin rights, “virtual on virtual”) and storage/versioning needs; ~75 survey respondents influenced 0.7→1.0 priorities.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 0:36Kubeflow Summit
- 6:10Why Kubeflow
- 6:55FinTech contributions
- 8:08Implementing Kubeflow
- 9:37Why erecto
- 12:31JPMorgan Chase
- 19:41US Bank
- 25:09Google
- 26:11Community Surveys
- 28:16Who is using Kubeflow
- 30:40Kubeflow components
- 32:03Kubeflow usage
- 34:07Key adoption blockers
- 35:49Outro
related talks
31mKubeflow - Data Science on Steroids
36mIntroducing Kubeflow (w. Special Guests Tensorflow and Apache Spark)
42mKubeflow From the End User’s Perspective: The Good, The Bad, and The Ugly
21mIntroducing Kubeflow (w. Special Guests Tensorflow and Apache Spark)
36mIntroducing Kubeflow (w. Special Guests Tensorflow and Apache Spark)
31mTransforming Data Science at PepsiCo: The Kubeflow Revolution
35mKubeflow: Data Scientist ready?
31mIntroduction to Using and Use Cases of KubeFlow