A Hitchhiker Guide to MLOps
summary
Claims that monitoring the ML lifecycle is practical end to end using only open source, and demonstrates it with MLflow for run/metric/artifact tracking and Grafana for dashboards and alerts. Shows concrete wins: auto‑logging models, saving datasets and configs as artifacts, setting thresholds that notify Slack, and cites 5 million downloads and 300 contributors as evidence of maturity. Builds a working pipeline that logs data drift and model metrics and reports that it is simple to deploy and useful in production.
AI-written summary · about
chapters (25)
- 0:00Introduction
- 2:10Do I need to care about it
- 4:20The main question
- 8:30MLOps definition
- 11:10Great expectations
- 12:00What is Flow
- 15:25Flow Setup
- 17:55Log Metrics
- 18:45Distance Metrics
- 19:30Context Managers
- 20:25Log Tags
- 20:50Log Artifact
- 21:55Performance
- 22:50Autolog
- 24:35Continuous Monitoring
- 25:30Grafana
- 26:05Panels
- 26:50Alerts
- 27:40Displaying Metrics
- 28:45Do I need to monitor ML
- 29:00My experience
- 30:30Education
- 31:35Getting started
- 32:05Questions
- 32:45Wrapup
related talks
24mA Hitchhiker's Guide to the MLOps Experience
41mMLOps in 30 Minutes
43mNavigating the MLOps Journey: Key Considerations for Successful Implementations
34mMLOPS 101: Tips, Tricks and Best Practices
13mMLOps at Scale
01h 13mPractical MLOps- Do we need all the things
13mMLOps: From AI Research to Production-Ready Systems
53mExperience with an end-to-end MLOps pipeline: Practical insights and 3D architectural perspectives