Professional Knowledge

Knowledge map

A connected map of the concepts and engineering practices needed to turn probabilistic models into useful, observable, and dependable systems.

LLM Systems

Tokens, context, inference, prompting, structured outputs, and model selection.

Explore →

Retrieval Systems

Indexing, ranking, context assembly, grounding, and freshness.

Explore →

Agent Systems

Tools, state, control loops, permissions, and bounded autonomy.

Explore →

Evaluation

Task suites, graders, experiments, error analysis, and release gates.

Explore →

Reliability

Failure containment, retries, fallbacks, budgets, and safe degradation.

Explore →

Observability

Traces, semantic signals, cost, latency, drift, and feedback loops.

Explore →

A system, not a stack

AI engineering is the work of shaping behavior across model, data, orchestration, interface, and operations. Local improvements can create global regressions: more context can reduce attention, more tools can expand the failure surface, and more retries can quietly multiply cost. The map is designed to reveal those connections.

Reading pattern

Every deep dive follows the same durable frame: Why → Mental Model → Core Concepts → How It Works → In Harness → Engineering → Trade-offs → Common Mistakes → Practice → Sources.