04 Entries · Archive
Field Notes.
Dispatches from the slower work of shipping AI systems — the AWS plumbing under them and the people side of architecting at the C-level.
Why your RAG retrieval is bad (and it's probably not the embeddings)
After shipping half a dozen RAG systems into production, I keep seeing the same retrieval failures. The model is rarely the problem — chunking and metadata are.
Tool-calling is just a remote procedure call with anxiety
How I think about designing tools for agents — naming, scoping, idempotency, and why your agent keeps looping.
A boring AWS architecture for a not-boring AI product
The reference architecture I've used to ship three production AI systems. ECS, Lambda, SQS — nothing fancy, and that's the point.
Notes from architecting at the C-level
What changes when the technical decisions you make end up in a board deck three weeks later.