18. AI Engineering¶
The discipline of building reliable systems around models. Children are practice areas: input-side (prompt/context/RAG engineering, tool/structured-output design), process (spec-driven and eval-driven development), and operations (observability, monitoring, cost/latency optimization, deployment). The unifying premise: the model is a fixed, stochastic component you can't retrain at will, so the engineering happens in everything surrounding it.
Children¶
- prompt engineering
- context engineering
- RAG engineering
- agent workflow design
- tool design
- structured output design
- spec-driven development
- eval-driven development
- observability
- monitoring
- cost optimization
- latency optimization
- deployment
Related¶
- Inference — the control surface engineers tune
- RAG — RAG engineering
- Evaluation & Testing — eval-driven development
- Infrastructure & Runtime — deployment and serving