Introducing ModelOps: How to deploy and manage every type of model
summary
Shows a concrete pipeline that retrains and replaces a production XGBoost gradient boosting model (version 4) after an alert for misclassification threshold 0.28 is raised, packages artifacts from JupyterLab into a zip, registers to SAS Model Manager, pushes to Git, runs Tekton CI tests (code quality, dependency and sanity checks), builds a container image and deploys via Argo CD rolling update, then runs a manual score; misclassification drops from ~30% to below 28 so the replacement succeeds.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 0:45Agenda
- 3:03ModelOps Components
- 4:10Model Lifecycle Diagram
- 5:18Summary
- 7:53Use case
- 9:20Next steps
- 11:04Example
- 17:44Demo
related talks
23mAccelerating Model Deployment with MLOps
51mSpeeding up deployment of machine learning models
57mDevOps for Machine Learning
36mData Science Meets DevOps: Gitops with OpenShift
24mFrom Model to Production: Automating the ML Lifecycle with CI/CD
50mBuilding an MLOps strategy from the ground up
46mProduction Model Deployment
25mBuilding Scalable MLOps Pipelines That Actually Work