A Data Science Operationalization Framework
summary
A small prototype framework exposes models as a REST API and stores trained artifacts in a Redis backend, implemented in Python with Flask and compatible with scikit‑learn; it supports automatic periodic retraining on incoming streams and demos online linear regression that retrains after ten new points and visibly adapts to distribution shifts. The code is deployable on Cloud Foundry and Heroku, includes basic visualization and model/version tracking, and can hook to Spark for batch training — a working, open‑source proof of concept with clear deployment and scaling choices to evaluate.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:40Enterprise data science challenges
- 4:08Models die in PowerPoint
- 7:22Requirements for model ops
- 8:23Operationalization framework
- 13:02Live demo
- 18:29Model training visualization
- 21:02System integration
- 23:51Enterprise framework expansion
- 26:21Open questions
- 28:27Comparison with solutions
- 29:11Real-world use case
- 29:34Q&A
related talks
46mProduction Model Deployment
59mDevOps for Data Science
32mDeploying Python models to production
38mData Secrets from a Platform Engineer
33mHow you really get your data science models into production the cool way!
41mFrom Prototype to Production Building a Scalable MLOps Platform from a Reference Arch
34mData Products, From 0 to data science pipeline
27mDeveloping and Deploying Models at Scale — without Spark!