Better Code for Data Science
summary
Argues that pragmatic engineering habits — project layout, tests, packaging, dependency pinning, code review and reproducible environments — turn notebooks into production systems without making data scientists full-time engineers. Demonstrates Python tools: data classes, linters and formatters (autopep8 black), conda/pip environments (he reports 45 environments on one machine), logging and IDE debuggers, and a small spaCy NLP module for German/English that yields concise notebook outputs. Claims team skill gaps can be a gap of 10x in productivity and shows concrete refactor steps that actually work.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 3:18Programming skills matter
- 5:23Data science personas
- 7:11Code architecture and stack
- 13:32From spaghetti to maintainable
- 24:57Jupyter to production
- 33:50Debugging and testing
- 38:05Key takeaways
related talks
28mThe Coding Portion of Data Science
39mRight Code, Right Place, Right Time
27mNotebook To Production
25mOvercoming 5 Hurdles to Using Jupyter Notebooks for Data Science, by the JetBrai...
30mA data scientist's guide to code reviews
37mPatterns for Collaboration Between Data Scientists and Software Engineers
42mSimplifying Data Analysis with GitHub Codespaces, Jupyter Notebooks & Open AI
26mWorking with Data Scientists and why I left my job (not)