The 20-Year Port: from PhD project to polished professional Python packaging
summary
A twenty‑year MATLAB geometrics toolbox becomes a production Python package by porting code, tests and docs, using coding agents (Google anti‑gravity harness, G Gemini, CL Claude, GitHub Copilot) to translate and implement improved algorithms, convert examples to doctests and automated test harness, set up CI/CD, type checks, ReadTheDocs, and PyPI releases; careful verification catches model hallucinations (wrong formulas), so adopt defense in depth with pre‑commit/pre‑push tests, randomized corner‑case tests, and a maintenance guide.
AI-written summary · about
chapters (12)
- 0:0020-year port project intro
- 1:33MATLAB to Python in 2026
- 2:57PhD origins and MATLAB code
- 6:12Delay in Python porting
- 7:57Modernizing MATLAB toolbox
- 9:23Coding agents for porting
- 11:50AI-assisted code challenges
- 13:37Robust software defense
- 16:36C++ library with Python front-end
- 18:26Porting takeaways
- 21:38Documentation and visualization
- 23:27Closing GitHub issue
related talks
22mPython for science, side projects and stuff!
20mCustom Python Applications in Neuroscience
55mAccelerated Python: The Community and Ecosystem
32mTips and tricks: data science prototype into production
29mBuilding scalable and robust Data Science Solutions
17mR & Python playing nice, in Production
43mLightning Talks
25mPipeline-level differentiable programming for the real world