End-to-End ML
summary
Cloudera shows an end-to-end churn pipeline that claims only 20% of models reach production, built on Cloudera Data Platform and CML using Kubernetes, Spark/Hive for ingestion, pandas and scikit‑learn for modeling, and LIME for explainability. The demo trains a logistic regression + LIME explained model, saves it as a pickle, and deploys it as a REST endpoint with an app that calls the model while a simulator makes 1,000 calls and injects errors to force accuracy decline and trigger retraining. The deployment, monitoring and versioning workflow works in the demo.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 0:42Design Decisions
- 1:37Machine Learning Failure
- 9:50Prototype Catalog
- 11:37interpretability
- 14:26Data Science Workbench
- 17:37Python Session
- 21:32Hive Table
- 23:02Data Visualization
- 24:37Model Build
- 28:27Model Training
- 30:42Model Deployment
- 33:12Model Simulation
- 34:17Model Interaction
- 42:32Summary
- 44:12Questions
- 48:12Data Engineering
related talks
27mA Look Inside Applied Machine Learning
24mFrom Model to Production: Automating the ML Lifecycle with CI/CD
31mMLflow: Platform for Complete Machine Learning Lifecycle
35mAn End-to-End Machine Learning Ecosystem in a Quarter
25mContinuous ML Applications in Production
10mML-based Log Analysis for Faster Debugging
47mMachine Learning in Production: An Intro to MLOps.
18mBeyond Inference: Bringing ML into Production