Good retrieval comes before a good answer
A practical walkthrough of a document-search pipeline, and how to find out which step is failing.
EXPLANATIONS & EXPERIMENTS
Writing.
Machine learning, explained one idea at a time.
Start here
Transformers, from tokens to attention
Follow a sequence through self-attention, with a diagram, the key equation, and a small PyTorch example.
Read the article
More articles
03 in the notebookA validation score you can trust
Data splits, leakage, and a small pipeline that keeps preprocessing inside the training boundary.
Example articles for the portfolio, with original diagrams and links to the source material.