Interview
summary
Description of an open source platform that runs ML on Kubernetes called Kubeflow, positioned as a shortcut to avoid building startup-grade infra from scratch, with a concrete recent addition: Kubeflow pipelines announced last week to handle data pipelines. Claims include applying Kubernetes architecture best practices to ML, reuse of standard software dev patterns, and active work to integrate other technologies and solve specific orchestration problems. Also notes early involvement via Scala and Spark and lessons carried into design.
AI-written summary · about
related talks
35mOrganizing data engineering around Kubeflow
35mBringing Your Data Pipeline into The Machine Learning Era
33mDiscover How to Create Your Own Metadata-Driven ML Platform from Scratch
18mAutomate the boring ML stuff with pipelines
28mFuture of Data Platforms: making use-case driven...
35mSimplification of Data Pipelines by Containerization
41mLeveraging open-source and 3rd party tools for ML on AWS
46mScaling ML Pipelines with Feast, Ray and Kubeflow