Reference explainers for ML fundamentals: math, architectures, and implementations from scratch.https://www.vitorsousa.com/https://www.vitorsousa.com/favicon.svghttps://www.vitorsousa.com/https://www.vitorsousa.com/foundations/building-a-transformer/https://www.vitorsousa.com/foundations/building-a-transformer/Layer normalization, residual connections, feed-forward networks, and the full decoder-only transformer assembled piece by piece with tested PyTorch code.Fri, 24 Apr 2026 19:00:00 GMThttps://www.vitorsousa.com/foundations/positional-encoding/https://www.vitorsousa.com/foundations/positional-encoding/Sinusoidal, learned, and rotary position embeddings — the math of why position information is necessary and how each approach encodes it.Thu, 02 Apr 2026 19:00:00 GMThttps://www.vitorsousa.com/foundations/attention-from-scratch/https://www.vitorsousa.com/foundations/attention-from-scratch/The complete guide to attention mechanisms — from the matrix math of scaled dot-product attention through multi-head projections to a tested PyTorch implementation. Every equation derived, every dimension explained.Wed, 01 Apr 2026 19:00:00 GMT