What the AI-era software engineer actually does: reading unfamiliar codebases fast, debugging systems you did not write, and using AI as a tool without trusting it blindly. Concept pages, no code.
How LLM-powered features fail in practice — prompt and config bugs, retrieval mismatches, retry and error handling. Concept pages plus real debugging problems.
Agents that loop, token budgets, caching, and streaming. Medium debugging problems drawn from real production incidents.
The subtle, high-signal failures interviewers love: silent data drift, tool-composition hallucinations, vector-store mismatches. Company-tagged problems with golden-answer comparison.
Company-tagged problems only. Required 15-minute simulator capstone. Completion generates a shareable certificate.