Longer notes on systems, product craft, and building under real constraints. Search the archive or open a post below.
20 articles
A practical guide to traditional word embeddings, contextual embeddings, why transformers rely on them, and how self-attention calculates them step by step.
A practical guide to the normal distribution, standard normal variates, z-scores, skewness, CDF, and why Gaussian assumptions matter in real data science workflows.
A detailed, beginner-friendly explanation of masked self-attention in the transformer decoder, why inference is autoregressive, how teacher forcing enables parallel training, and how the causal mask prevents cheating.