TwinQuery AI — RAG over city policy

Policy lived in hundreds of long PDFs. Staff and residents could not find what applied to them, and a generic chatbot would make things up.

• ✦Document ingestion with layout-aware chunking and metadata
• ✦Multilingual embeddings and a vector index with hybrid search
• ✦Answer generation grounded in retrieved passages, with citations
• ✦An evaluation set to measure answer quality before each release

Answers people can check, in their own language, in seconds instead of an afternoon of searching.