Real-Estate RAG Assistant & Chatbot

Situation

A real-estate company handled a large volume of property documents — listings, technical diagnostics, legal records. Staff repeatedly searched and cross-referenced these documents to answer questions from clients and internal teams.

Task

The client wanted a natural-language assistant that could answer questions about the corpus quickly and with verifiable sources, replacing manual lookup work.

Action

I designed and delivered a retrieval-augmented generation (RAG) system with a chatbot front end:

• Indexing: property documents parsed, chunked, embedded, and stored in Qdrant.
• Retrieval: top-k vector search with metadata filters to scope answers to the relevant document types and dates.
• Chatbot: a conversational interface exposing the retrieval pipeline, with answers citing their source chunks.

Result

• Staff answered document questions in conversation instead of through manual searches.
• Citation-backed answers made responses verifiable — important for documents with legal weight.
• The retrieval pattern was reusable across the client's other corpora.

Confidentiality note

Client and property details stay general here, following the same posture as my other regulated case studies.

Stack highlights

• LangChain for the retrieval pipeline
• Qdrant as the vector store
• Python for all ingestion and query-time logic