Jędrzej Maczan

Hi!

I like computers. Nowadays I focus on deep learning compilers, which means working on PyTorch internals, coding in Python, C++, CUDA and WebGPU (WGSL). Sometimes I code in Rust and TypeScript

Night volunteer job: ML compiler engineering @ PyTorch and building torch-webgpu - WebGPU backend for PyTorch (compiler and runtime)
Day job: Applying LLMs to things
GitHub: @jmaczan
HackTheBox: @7563687575 (inactive, had fun some time ago)
LinkedIn: click
X/Twitter: @jedmaczan
Email: jedrzej@maczan.pl
Anita's paintings: anitamaczan.pl, instagram: @anita_maczan

writing

Use 20x less peak RAM with dp_knapsack_sliding_hirschberg, a new activation memory budget solver in PyTorch, 2025 GitHub
Self-contained handwritten digit recognizer, Paged Out!, Issue #7, page 11, 2025 GitHub
A primer on Differentiable Architecture Search, Paged Out!, Issue #6, page 5, 2025 GitHub
GPT in PyTorch, Paged Out!, Issue #5, page 6, 2024 GitHub
Building automated machine learning with type inference, Paged Out!, Issue #4, page 4, 2024 GitHub

 . 
  .
...