ai
What a RAG evaluation harness actually needs to measure
Retrieval quality and answer quality fail in different ways, and a single score hides both. Measure the two stages separately, on a set you wrote by hand.
Engineering notes
Writing about the work
Architecture decisions, tradeoffs, and things that broke — written by the people who build.
- ai 2 articles
- cloud 1 article
- delivery 1 article
- evaluation 1 article
- terraform 1 article
cloud
Terraform landing zones: the five decisions you can't undo later
Account structure, naming, state layout, network topology and identity. Everything else in a landing zone is reversible; these five are not, so decide them first.
ai
Why your first ML project should be boring
The projects that survive contact with production are the ones whose failure mode is obvious. Pick a problem where being wrong is cheap and visible.