Scaling MLOps on Kubernetes with Amazon SageMaker Operators
summary
Kubernetes CRDs call out training and hosting to SageMaker Operators while Argo workflows handle GitOps CI/CD, Jenkins builds container images, and logs stream to Splunk. Intuit’s platform runs over 400 production models, serves over 15 billion+ features/day from ~8,000 features, injects VPC, IAM role, encryption and autoscaling policies automatically, and triggers retraining via Model Monitoring and CloudWatch alerts. Demo shows a Git commit → Argo → SageMaker training job that completes successfully and returns artifacts to S3.
AI-written summary · about
chapters (6)
- 0:00Introduction and speakers
- 1:04MLOps and SageMaker overview
- 3:11SageMaker Operators benefits
- 10:31Intuit’s ML platform scaling
- 16:38Demo of Kubernetes integration
- 27:27Closing and acknowledgments
related talks
34mImplement MLOps practices with Amazon SageMaker
25mBuilding MLOps POCs and Sandbox Environments Using k3s and Argo
26mManaging ML Models @ Scale
36mImplementing MLOps practices on AWS
37mStreamlining Machine Learning Operations
34mImplement MLOps practices with Amazon SageMaker
12mLeveraging AWS and Kubernetes for Scalable AI/ML Workloads in the Cloud
33mBuild, train, and deploy ML models using Amazon SageMaker