AI Pharma Document Q&A Tool

I need a lightweight web application that lets a user drag-and-drop a pharmaceutical PDF—whether it’s a drug label, clinical study, SOP, regulatory guideline, or research paper—and then type a question in English to receive an answer drawn from that document. The core search layer must blend straightforward keyword matching with a vector-based approach so the system can return contextually relevant passages rather than just isolated hits.

Here is what will make the build a success:
• Upload interface restricted to PDF for this first release.
• Dual search pipeline: classic keyword lookup plus contextual retrieval using embeddings so follow-up questions such as “What were the primary endpoints?” surface the right section immediately.
• Chat-style UI that echoes the question, displays a concise answer, and highlights the cited sentences inside the document.
• English is the only language requirement at this stage.
• Clean, self-contained code (Python, Node, or comparable stack), clear setup instructions, and a short README explaining how to replace or scale the embedding model later if we decide to add full NLP.

The deliverable is a functioning proof of concept I can demo to non-technical stakeholders: deployable on a standard cloud VM or run locally with a single command, no hidden dependencies, and no external data leakage.