Inspiration

In every company, knowledge is scattered - in PDFs, Notion, Slack, Google Docs. New employees take weeks to find answers, and existing teams keep answering the same questions. We wanted to build an Enterprise Brain that actually understands your data and answers with proof, not hallucinations.

What it does

*EnterpriseIQ (Clever) is a grounded RAG-based Knowledge Management System.

Ingest Anything:Upload PDFs, docs, wikis - it auto-chunks and embeds them.
Grounded Answers: It never hallucinates. Every answer comes with the exact source document and page number.
Built-in Grounded RAG Engine: No external LLM dependency for retrieval, our engine does semantic search + re-ranking.
Public Web Service:Live, HTTPS-secured, responsive on desktop and mobile for judges to test.

Ask it "What is our refund policy?" - it answers from YOUR docs, not the internet.

How we built it

Backend: Python, FastAPI, Vector DB (FAISS/Chroma)
RAG Pipeline:Sentence-Transformers for embeddings, custom chunking, grounded prompting
Frontend:Clean, responsive chat UI
Deployment: Dockerized and deployed on Render as a public web service with Docker build
Auth & Storage: Persistent document store with source tracing

Challenges we ran into

Preventing hallucination - solved by forcing RAG to answer only if source score > threshold.
Docker build on Render failing due to large model size - optimized with smaller embedding models.
GitHub auth issues with multiple accounts (403 permission) during deployment.
Handling large PDFs without losing context in chunks.

Accomplishments that we're proud of

Built a fully working grounded RAG that cites sources - not just a ChatGPT wrapper.
Successfully deployed as a live public URL that judges can open instantly.
Response time <2s even on free tier.
Works on both mobile and desktop.