Engineering notes

Shipping AI systems that actually work.

Practical writing on evaluation, retrieval and ML infrastructure from engineers who carry the pager.

Start here

You cannot improve what you only feel. A lightweight evaluation loop you can set up in an afternoon, and why it pays for itself in the first week.

Maya OkaforOct 12, 20263 min read

Most retrieval-augmented generation failures are retrieval failures. Twelve checks to run before you blame the model.

Maya Okafor & Daniel ReyesOct 6, 20263 min read

A retrospective on launching an AI writing assistant: what we got right, what surprised us, and what we would do differently.

Maya Okafor & Daniel ReyesSep 7, 20262 min read

You cannot improve what you only feel. A lightweight evaluation loop you can set up in an afternoon, and why it pays for itself in the first week.

Maya OkaforOct 12, 20263 min read

Most retrieval-augmented generation failures are retrieval failures. Twelve checks to run before you blame the model.

Maya Okafor & Daniel ReyesOct 6, 20263 min read

Per-token prices keep falling, yet our bill kept rising. Where the tokens were really going, and the five changes that cut spend by 61%.

Daniel ReyesSep 29, 20262 min read

Getting reliable JSON out of a language model is a solved problem — if you combine schemas, validation and a sensible retry policy.

Maya OkaforSep 22, 20262 min read

Four GPUs, nine engineers and a lot of idle time. How a simple queue and a few house rules doubled our utilisation.

Daniel ReyesSep 14, 20262 min read

A retrospective on launching an AI writing assistant: what we got right, what surprised us, and what we would do differently.

Maya Okafor & Daniel ReyesSep 7, 20262 min read