Hands-on understanding of how large language models work, how to evaluate them, and how to compose them in real systems. Covers transformer fundamentals, contex
LLM skill is working mastery of the engine: model selection and routing, API depth (streaming, tools, structured output), context and cost management, and the behavioral intuition, failure modes, sampling, capability edges, that turns model access into product reliability.
The center of gravity: nearly every AI role lists LLM fluency, and depth, behavioral intuition plus cost engineering, separates senior offers from keyword matches.
Any frontier API deeply (Claude is an excellent start): skills transfer almost completely. What compounds is task-fit judgment and evaluation habit, not provider trivia.
Failure-mode fluency and trade-off judgment: how you'd cut hallucinations, manage context budgets, choose models per task, and prove quality. War stories beat specs.