Practical MLOps for better models
summary
Claim: vetiver actually versions, deploys, and monitors models end to end with concrete tooling. It uses vetiver model objects pinned into pins boards, stores a small training prototype for debugging, and generates a quarto model card with metrics like MAE/MSE. APIs spin up with a few lines using FastAPI, app.py and Dockerfile are auto-generated for cloud or on‑prem, metrics compute over rolling windows and pin results, single‑dispatch limits dependency bloat, and the examples run locally and deploy successfully.
AI-written summary · about
chapters (12)
- 0:00MLOps introduction
- 2:12Bringing models to life
- 6:19What is Vetiver and why it matters
- 8:25Versioning models
- 11:00Model cards for ethics
- 14:11Deploying with Vetiver API
- 16:43Monitoring models in production
- 18:50Why Vetiver excels
- 20:18Closing remarks
- 21:41Q&A: Vetiver dependencies
- 23:47Q&A: Decentralized vs centralized
- 24:59Q&A: Vetiver vs MLflow
