AI/ML / 2024 / Maintained
AI Document Q&A Agent
Document ingestion and retrieval system that answers questions with citations linked to retrieved source chunks. Runs in a deterministic mock mode so reviewers can exercise the full pipeline without paid LLM keys.
Role
Designer and engineer (solo)
Disciplines
AI Integration / Backend Systems / Full-Stack
Status
Maintained
What this system solves
Generic chatbots hallucinate and offer no audit trail back to the source material — unusable for any document-grounded workflow that needs verifiable answers.
System overview
FastAPI + React product: upload panel, document library, question workbench, answer panel, and citation list. Pipeline covers ingestion, bounded chunking, lexical retrieval, citation builder, and a swappable mock/OpenAI answer generator behind a stable provider boundary.
Limitations
- Lexical retrieval only — no embeddings or vector index in the current build.
- In-process document store; not designed for high concurrency or large corpora.
- Mock provider is for demonstrability; real synthesis requires a configured OpenAI key.
Validation and results
Reproducible portfolio MVP that demonstrates real RAG product thinking — citations, evaluation surface, and a swap-in path to production LLMs.