1
Overview
What belongs in this track once it's written.

Step 5 of the AI Engineer roadmap: the frameworks used to build LLM applications, plus the framework-agnostic patterns interviewers actually probe (why this framework, when to use it, its limits, how you'd run it in production).

  • LangChain
  • LangGraph
  • LlamaIndex
  • Tool calling
  • Structured outputs
  • Memory (short-term, long-term, and framework-managed)
  • What happens when the agent selects the wrong tool? (detection, recovery, guardrails)
MCP (Model Context Protocol) has its own dedicated track — see that section instead of duplicating it here.
Fill this page in question-by-question the same way the other tracks are written, then this callout can go.