Productionalizing Data Science Models
summary
Claim: notebooks are not production code and must become maintained artifacts. Practical steps: convert notebooks to pure Python and put code in git; save static outputs and experimental datasets in shared fault tolerant storage; separate into files: data cleaning, model training, prediction, plus modular functions; add tests, automated metric collection and deployment thresholds that reject models which fail benchmarks. Also consider integrating models into ETL or providing stable APIs.
AI-written summary · about
related talks
39mNotebook To Production
28mThe Coding Portion of Data Science
35mGetting Machine Learning Models Ready For Production Using Python
20mBetter Care and Feeding of Machine Learning Models
29mBuilding scalable and robust Data Science Solutions
17mProductionizing Data Science Workloads
40mThe data scientists are gone. how do I deploy their work to prod?
29mData Science Best Practices : From Proof of Concepts to Production