The Discrete Fourier Transform: From Math to FFT
Working through the DFT from its definition to the Fast Fourier Transform — why the naïve O(n²) algorithm collapses to O(n log n), and what that means in practice.
Notes on Gradient Descent
Working through the math of gradient descent — how the loss surface shapes training, and why small implementation details matter more than I expected.
Attention Mechanisms in Transformers
Working through scaled dot-product attention from scratch — the query/key/value framework, why it works, and the one scaling detail I kept overlooking.