Skip to content

LLM Engineering

Context engineering: deciding what the model gets to see.

LLM APIssystem promptsfew-shotstructured outputfunction callingcontext engineeringsummarization

Lessons

  1. 21.1Prompts and message rolesA prompt isn't one big string. It's a short conversation with roles, and treating it like real code saves you from silent breakages. 30 min Prompt injection lab Eval playground
  2. 21.2Structured output: making LLMs return data your code can trustModels write text. Your code needs data. Here's how to get answers back in a fixed shape you can actually trust. 30 min Next-token sampler
  3. 21.3Context engineering: deciding what the model gets to seeThe model only knows what you show it on each call. Choosing what goes in, and what stays out, is a skill of its own. 30 min Context window builder

On the job: try it on your own project

Find one place where your product calls an LLM, or one you want to add. Ask Claude Code to show you the full prompt as sent: system prompt, examples, context and output format. Look for vague instructions and missing structure. Write an improved prompt spec with a structured output schema, then test old and new on five real inputs and record which wins.

Can you do this now?

0/4. Anything you can't tick yet: reopen that lesson's practice ladder.