Feb 21, 2026

Business Requirement Document (BRD) Generation Agent

api fastapi gemini rag faiss

Organizations conduct countless meetings where critical requirements, decisions, and 
timelines are discussed but remain buried in lengthy, unstructured conversations. Manually 
converting these discussions into formal Business Requirement Documents is slow and 
error-prone. This leads to miscommunication, conflicting expectations, and project delays, 
ultimately increasing costs and reducing overall execution efficiency.

Our Idea: ReqMind AI
An intelligent BRD Generation Agent powered by Turgon AI that transforms unstructured meeting conversations into structured, 
explainable Business Requirement Document.
Our Approach
❖ Transcript Ingestion – Accepts AMI meeting transcripts as input
❖ Intelligent Noise Filtering – Removes filler dialogue and irrelevant discussion
❖ Structured Extraction – Identifies stakeholders, functional & non-functional requirements, timelines, and 
decisions
❖ Conflict Detection – Flags contradictory statements within discussions
❖ Automated BRD Generation – Maps structured output into a professional BRD template
❖ Traceability Matrix – Links each requirement to speaker & timestamp

This build was uploaded as a hackathon project

Hackathon

HackFest 2.0

View All Projects
Give a star to encourage!Discussion
Start a new conversation!
Login to join the discussion

More Builds by Aqeedat Insha

ml ai api agora hackathon
Updates
  • Made backend into Fast API and deployed it on Google Cloud
    Sunday, Feb 22nd, 2026
  • done with frontend, backend and API integration
    Saturday, Feb 21st, 2026
  • 🚀 Development Phase Update — ReqMind AI Team: CodeBlooded | HackFest 2.0 We are excited to share the latest progress on ReqMind AI, our RAG-powered Business Requirements Document (BRD) generation system. 🔹 Phase 1: Architecture & Backend Foundation — Completed Designed a 9-stage RAG-based pipeline for structured requirement extraction. Built backend using FastAPI, deployed on Google Cloud Run. Integrated Gemini 2.0 Flash as the LLM layer. Implemented FAISS vector indexing with MiniLM embeddings for semantic retrieval. Added rule-based preprocessing (speaker segmentation, filler removal, noise filtering). ✅ Backend is fully functional and live. 🔹 Phase 2: RAG & Intelligent Extraction — Completed Implemented semantic chunking with overlap for contextual preservation. Built multi-query retrieval strategy to improve extraction accuracy. Developed: Functional Requirement extraction Non-Functional Requirement extraction Stakeholder sentiment analysis Conflict detection Timeline & decision extraction Added confidence scoring with HITL (Human-in-the-Loop) flagging for low-certainty outputs. ✅ System now produces structured JSON and traceable outputs. 🔹 Phase 3: BRD Generation & Formatting — Completed Created a fixed BRD template engine. Automated section-wise population: Executive Summary Project Overview Stakeholders Requirements Conflict Log Traceability Matrix Enabled iterative BRD editing via AI prompt refinement. Implemented TXT & DOCX export functionality. ✅ Production-ready BRD generation achieved. 🔹 Phase 4: Frontend & UX — Completed Built responsive HTML-based frontend. Implemented: Live pipeline provenance display Executive summary viewer Requirement cards with traceability Interactive edit feature Download & matrix view Connected frontend to deployed Cloud backend via API. ✅ End-to-end working product available. 🔹 Phase 5: Evaluation & Dataset Integration — Completed Integrated AMI Meeting Corpus via HuggingFace REST API. Implemented fallback mechanism for robustness. Added pipeline introspection endpoint for transparency. ✅ System validated on real-world meeting transcripts. 📊 Current Status Backend: Live (Cloud Run) Frontend: Fully integrated RAG Retrieval: Functional BRD Generation: Stable Iterative Editing: Working Deployment: Production-ready demo build
    Sunday, Feb 22nd, 2026