Back to work
AI / ML Projects2024–

AI Learning Experiments

Small end-to-end reproductions — from Titanic to spike sorting — written to internalize the algorithms, not just cite them.

NumPyPyTorchNotebooks

Habit, not portfolio

A growing folder of small end-to-end reproductions I wrote to actually understand the algorithms — not to fill a resume.

Recent notebooks

  • A minimal Transformer from scratch, trained on tiny text
  • Spike sorting on public electrophysiology data
  • A tiny diffusion model on 32x32 images

What I take away each time

One page of notes on what surprised me. Over time, those pages become the actual curriculum.