AI Engineering

AI engineering here means the work around a model, not the work of training one. The pages in this branch are about pipelines that retrieve, checks that tell you whether an answer got better, and agents that call tools under a contract. The model is an input. The system you can operate is the subject.

The linear path places this branch after architecture. You can still open it directly. The declared prerequisites are a backend that can hold a request boundary, and enough distributed-systems context to treat a tool call as a remote call that can fail, retry, and arrive twice. You do not have to read every database page first. If a later post needs a queue or an outbox, it says so.

Two sub-branches carry the published work. LLM fundamentals holds the retrieval pipeline and the evaluation notes, in that order, because measuring a system you cannot describe is how teams fool themselves. Agentic workflows comes after, and it points back at messaging and at those fundamentals. I am not going to pretend an agent loop is a beginner topic.

Maturity defaults to budding on purpose. These pages change when the platform changes, and a confident badge on a moving target would be a lie. When a note is still rough I will mark it seedling. Read the badge before you copy a design into a production service.