Generative AI Application Development

Build a source-grounded generative feature with evaluation checks and reliability notes.

Problem statement · Illustrative

Users need grounded answers from a curated knowledge set, with evaluation that catches unsupported claims.

Primary user workflow · Illustrative

Ask a question → retrieve sources → generate grounded answer → run evaluation checks → return answer or safe refusal.

Functional scope · Illustrative

LLM application structure, structured outputs, basic retrieval, evaluation harness, and responsible data handling notes.

Technical components · Illustrative

Application service, model API integration, retrieval layer, evaluation cases, and documentation.

Data requirements · Illustrative

Fixed illustrative corpus. Production proprietary data is not included.

Testing expectations · Illustrative

Grounded answer case, missing-evidence case, and at least one malformed input or dependency failure.

Known limitations · Illustrative

An MVP with evaluation is not equivalent to a fully production-hardened AI system.

What counts as completion · Illustrative

Working grounded feature, evaluation notes, reliability write-up, and demonstration of failure handling.

Example or assigned project work—not a published student achievement or employment guarantee. Student work is not assumed for commercial use without appropriate agreements. Sections marked illustrative are representative examples, not a fixed commercial brief.