Pybites Podcast #210: Codeflash and continuous Python performance with Saurabh Misra
Jan 5, 2026
In this discussion, Saurabh Misra, founder of CodeFlash and a software engineer with a rich background at NVIDIA, reveals how to make Python performance optimization part of everyday coding practices. He shares insights on using LLMs to automate performance checks, profiling techniques, and real-world examples like improving Bokeh and Pydantic. Saurabh explains the importance of maintaining code behavior while delivering faster execution, and offers practical advice for integrating performance into team workflows.
AI Snips
Chapters
Transcript
Episode notes
Bokeh Real-World Optimization
- Saurabh and CodeFlash worked on optimizing Bokeh and are merging improvements upstream.
- Users of Bokeh should see performance gains once those PRs land.
Automating Expert Performance Workflows
- CodeFlash automates the steps a performance engineer would take to find faster code variants.
- It verifies behavior and measures speed to propose minimal, trustworthy diffs for maintainers.
Speedups In Open Source And Industry
- CodeFlash produced notable speedups for Pydantic and for computer vision/NLP pipelines in industry.
- These enterprise wins show optimization brings latency and cost benefits at scale.