About

CPython, asyncio, and chasing runtime performance

I’m Timofei Ivankov, a Python backend engineer who tends to end up where systems hit their limits — making ML inference faster, keeping async services stable under load, and reading the interpreter’s source instead of guessing how it behaves.

I contribute to CPython, mostly around asyncio and the free-threaded build — fixing data races, resource leaks, and correctness bugs, some of which landed in 3.14 and 3.15. You can see the work here.

This blog is where I write up the bugs I chase: what the symptom looked like, how I tracked it down, what was actually going on underneath, and what the fix was. Mostly CPython internals, concurrency, and runtime performance.

What I work on

  • CPython contributions — asyncio, free-threading, small JIT and runtime fixes.
  • Runtime performance — I run measurement studies on how the different ways of speeding up Python (JIT, native extensions, alternative runtimes, build-level optimizations) actually compare when measured under one protocol.
  • Backend & ML infrastructure — production services around LLMs and in-house models: serving, inference optimization, and keeping async systems stable under load.

I also write in Russian on Habr about distributed systems, fault tolerance, and highload.

Elsewhere

If you want to talk about CPython internals, async, or runtime performance, reach out.