The Problem
Logistics and legal teams sit on piles of scanned contracts and paperwork that are painful to search, validate, and export into structured formats.
What I Built
A multi-agent document intelligence platform that ingests contracts and scanned logistics documents, extracts structured fields via OCR + LLM pipelines, and exports validation-ready datasets.
How It Works
PaddleOCR handles document digitization. Ollama-powered LLM extraction maps raw text into schema-defined fields. A Next.js interface provides real-time validation and Excel export. The full stack runs containerized with Docker.
Outcome / Results
Finalist at the OLTEK Paper to Data Hackathon. Proved that OCR + local LLM extraction can produce export-ready business data from messy paper workflows.