In May 2025, I had the pleasure to give a presentation at the PyData Zurich meetup, Rust in the data science and machine learning stack.
After introducing the role of compiled languages in the Python ecosystem, the talk gave a short presentation of the Rust programming language (partly adapted from an earlier talk), before analyzing some success stories: ruff
, uv
, polars
, pydantic
and tokenizers
.
Then, a toy Byte Pair Encoding (BPE) tokenizer was used to illustrate the port of a Python implementation to Rust (including the creation of bindings) and the outcomes on performance. The code of the two implementations is available in a small Github repository.
See PDF (50 pages) or below for the slides, authored in Typst.