Accelerating Model Deployment with MLOps
summary
Claims that CI plus MLOps cut the 30+ day bottleneck for deploying models by automating the pipeline: a reproducible demo uses a repo with Dockerfile and .gitlab-ci.yml to run a Jupyter training notebook (XGBoost on sklearn toy data), four CI jobs (build container, run training, log metrics to merge request comments, store artifacts) and a secret scan. Models track to MLflow style experiment storage, deploy via tag releases, scale CPUs/GPUs, run scheduled CI for production, and trigger retrain on drift; everything works in the free GitLab runner setup.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 0:49Model deployment challenges
- 2:17DevOps in data science
- 3:22CI/CD for ML
- 5:29Demo setup
- 8:17Training in GitLab
- 11:28Model performance monitoring
- 13:40Deploying to production
- 15:49Automating production runs
- 16:34Advanced CI/CD features
- 18:00Resources
- 18:22Q&A
related talks
24mFrom Model to Production: Automating the ML Lifecycle with CI/CD
31m🤑 Paying Off Technical Debt in the Development of Machine Learning Apps
25mBuilding Scalable MLOps Pipelines That Actually Work
39mContinuous Delivery for Machine Learning — The Future of MLOps
32mMLOps at Renault: A Generic Pipeline for Scalable Deployment
10mFrom Model to Production: Building Scalable MLOps Pipelines
01h 17mMaking MLOps uncool again
08mMLOps in Healthcare: Scaling AI Model Deployment