3+1 Ways to Do MLOps with Kubernetes
summary
Kateryna Hrytsaienko examines production MLOps challenges and shows how Kubernetes supports model deployment, autoscaling, observability and cost/control at scale. The talk presents a practical “3+1” set of patterns and tools for running ML models on Kubernetes, covering deployment, scaling, monitoring and operational cost management.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:07What is MLOps and Kubernetes
- 5:183 ways to do MLOps with Kubernetes
- 7:27Platform way with Kubeflow
- 9:33Compute way with Ray
- 15:11Engine way with TGI
- 24:36Combining methods for production
- 31:19Free credits and resources
- 33:10Q&A
related talks
06mEvolving with Kubernetes: Embracing Model Ops
41mMLOps in 30 Minutes
47mMLOps tools to automate ML deployment
36mIs the MLOps tooling the same as the DevOps tooling?
47mMachine Learning in Production: An Intro to MLOps.
33mModel Development and Deployment Tools. An Introduction to MLOps.
10mFrom Model to Production: Building Scalable MLOps Pipelines
14mWhy Your DevOps Skills Are Your MLOps