AI / ML · case study
Contract Summary AI
Construction contracts are long, dense, and expensive to review. Contract Summary AI ingests a UK construction contract as a PDF and turns it into something a human can actually use: a structured summary, plus a chat assistant that answers questions grounded in the document itself.
The pipeline
The backend is a Flask REST API orchestrating LangChain. An uploaded PDF is parsed, chunked, and embedded into ChromaDB. When a user asks a question, the most relevant passages are retrieved and handed to OpenAI GPT-4o alongside the query, so every answer is grounded in the actual contract text rather than the model's imagination. The same retrieval machinery drives the structured summaries.
The interface
A responsive React frontend handles the full workflow: upload a contract, read the generated summary, then drop into a chat with the assistant for anything the summary didn't cover. The frontend is deployed on Vercel and talks to the Flask API.
What it demonstrates
End-to-end Retrieval-Augmented Generation shipped as a working product rather than a notebook demo: document processing, vector search, grounded generation, and a clean interface wrapped around all of it.