A transformer-based language model built entirely from scratch in NumPy — I wrote my own autograd engine, BPE tokenizer, multi-head attention, and Adam optimizer, with no ML framework involved. Built to genuinely understand how a language model works end to end, gradient math included. See it on GitHub.
I've been working through pwn.college's CTF and systems-security challenges for about a year, on and off — currently around halfway through the orange belt, with the goal of eventually finishing all of it.
This website is still under active development. I'd appreciate any suggestions — feel free to reach out on Discord, since that's where I'm most active.
