Skip to content
View horecdev's full-sized avatar
  • Warszawa
  • 13:08 (UTC +02:00)

Block or report horecdev

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
horecdev/README.md

Michał Horecki

Without rambling, I like building all AI type shit from the ground up.
Incoming CS freshman @ Warsaw University of Technology.

Three projects I find the coolest

  • GradCraft: 12,000 lines of C++/CUDA. A custom deep learning framework with a lazily-evaluated autograd engine, fused kernels (95x speedup on RMSNorm, 108x on Softmax Cross-Entropy), and custom memory pools. Trained a 180M parameter GPT at home on a 3090.
  • NumPy Audio Denoiser: A speech enhancement LSTM built entirely in raw NumPy. Hand-wrote the BPTT and STFT/ISTFT math.
  • Lyme CV Pipeline: A 3-stage TensorFlow vision pipeline for Lyme disease detection using multi-backbone feature ensembling.
  • ... and many more but this README.md is not my CV or a spreadsheet for all projects ever.

🔗 📄 Read my 1-page CV here


X (Twitter) | Email

Pinned Loading

  1. GradCraft GradCraft Public

    A C++ Autograd Engine written without ANY frameworks that can train a pretty big GPT locally. Built to maximize performance.

    C++

  2. LSTM-denoiser-from-scratch LSTM-denoiser-from-scratch Public

    Long Short-Term Memory built without high-level ML libraries

    Python