Considerations for producing the best AI and Machine Learning models
summary
Explains why models fail in production and how to fix it with concrete practices: build a feedback loop for ground truth even when outcomes take years (example: loan defaults can take ~10 years), choose batch scoring or real‑time deployment and use API endpoints across a Kubernetes cluster for scaling, instrument continuous monitoring to track model quality not just accuracy, measure business value with a cost matrix so each prediction maps to dollars, and use an end‑to‑end platform (Dataiku) to deploy without downtime and enable transparent review for fairness.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 1:25Bridging the gap
- 2:53How to judge a model
- 3:46Deployment options
- 4:59Best practices
- 5:37Endtoend platform
- 6:36Conclusion
related talks
16mOperationalizing AI in Finance: Real-Time Analytics
43mLearning the fundamental elements to build effective AI/ML products
30mFrom Black Box to a Known Quantity: How to Build Predictable, Reliable ML-based
59mWhy Great Models Fail: Lessons From 9 Years of Deploying ML Models
29mDemystifying Machine Learning in Production: Reasoning about a Large-Scale ML Platform
41mBeyond the Algorithm: What Makes Machine Learning Work?
48mPrinciples and Practices of Building Machine Learning Systems
28mLeading machine learning research journeys with happy (into-production) endings