How I Built an AI-Powered Document Intelligence Platform Using Next.js, FastAPI, RAG and MCP

Description and use

The platform automates document intake, extraction, validation, and human approval, reducing manual effort while maintaining accountability.

Problem

Manual processing of invoices, reports, contracts and other business documents is time‑consuming and difficult to manage efficiently.

Outcome

The platform automates document intake, extraction, validation, and human approval, reducing manual effort while maintaining accountability.

Workflow

• Upload PDF/JPG/PNG to backend
• Validate file type and size
• Store original in AWS S3 and metadata in PostgreSQL
• Classify document type and extract fields via LLM
• Validate extracted data against business rules
• Index structured data in ChromaDB for RAG
• Schedule review via MCP Google Calendar integration
• Manager reviews extracted info and approves or rejects
• Record approval status and audit trail

Evidence

Manual processing can become time-consuming and difficult to manage efficiently.

Source

The platform transforms uploaded documents into structured, searchable, and reviewable information.

Source

Upload → Classification → Information Extraction → Validation → Indexing → Human Review → Approval or Rejection

Source