Automating Python Code Performance
summary
Describes Codeflash, an open tool that automates performance improvements for Python by using language models to propose concrete code replacements, running automatic benchmarks, and generating or reusing tests to verify functional equivalence before and after changes. It instruments real executions to capture inputs, uses line-level profiling to target transformations, and opens pull requests when a validated suggestion improves runtime (example: from 137 ms to 530 µs, ~257× faster).
AI-written summary · about
chapters (8)
- 0:00Welcome
- 1:13Python isn't slow
- 3:23Manual optimization challenges
- 6:16Automating performance
- 8:02CFRAS demo
- 16:53Real optimizations in libraries
- 19:20Tool integration
- 22:54Q&A
related talks
10mSpeed Up Your Code With LLM's!
01mSpeed Up Your Code With LLM's!
40mCode Rot at Machine Speed: Preventing AI-Induced Technical Debt
28mInstant-feedback, instant-debugging Python coding
30mBreaking the Speed Limit: Fast Statistical Models with Python 3.14...
29mIs coding assistant as good as we thought
27mSpeed up Python with hardware accelerators
18mCoding Evals: From Code Snippets to Codebases