Build End-To-End MLOps Platform with Open Source DVC Ecosystem
summary
Step‑by‑step demo of an end‑to‑end MLOps stack that actually runs: git + DVC for data and model versioning (dvc init / dvc add / dvc push / dvc pull), dvc.yaml pipelines and dvc exp run for reproducible experiments, DVCLive metrics + VS Code extension for live metric/plot capture, CML + simple CI jobs to train, dvc push and register models via Git tags (GTO pattern), and a preview of dcx for scaling to billions of files — all shown with working commands and repository examples.
AI-written summary · about
chapters (14)
- 0:00MLOps and DVC intro
- 4:28Speaker and goals
- 11:40MLOps patterns explained
- 15:19Monolithic vs modular pipelines
- 29:19Designing MLOps platforms
- 35:21What is DVC
- 45:48DVC data versioning
- 1:06:48Switching data versions
- 1:17:54DVC pipelines and tracking
- 1:35:16Running experiments
- 1:44:08Model registry and CI/CD
- 1:53:54Monitoring data and models
- 1:58:17DVC Studio collaboration
- 2:07:41Q&A
related talks
29mDevOps for Data Science? - automate the boring stuff and leverage the OSS ecosystem
37mOpen Source Tools for ML Experiments Management
01h 32mHow to easily set up and version control your Machine Learning Pipelines
59mDevOps for Machine Learning Projects
42mApplications in ML Drug Discovery pipelines
31mDevOps for Science: using continuous integration
01h 17mMaking MLOps uncool again
27mLeveraging Git for ML Experiment Management