Project 01Live

Arco Papers

A paper trading business generates dozens of handwritten invoices every day. Arco Papers uses GPT-4o vision and a LangChain agent to photograph those documents and turn them into structured financial records in seconds, with zero manual data entry.

Launch App →

The problem

A family-run paper manufacturing and trading business in Islamabad was recording every purchase and sale invoice manually into spreadsheets. Handwritten documents in Urdu and English, multiple suppliers, daily volume. Reconciliation took hours and errors slipped through.

The solution

A Flutter mobile client lets staff photograph an invoice. A FastAPI backend passes the image to GPT-4o Vision via a LangChain agent, extracts line items and totals, validates the data, and writes structured records to Supabase. The web dashboard gives the owner a live view of inventory and financials.

How it works

GPT-4o Vision

Reads handwritten Urdu/English invoices, extracts line items, quantities, prices.

LangChain Agent

Orchestrates extraction, validation, and database writes in a single pipeline.

Live production

Running on real business data for a paper manufacturing and trading company.

Flutter client

Cross-platform: same codebase runs as web app and on Android.

Tech stack

Backend

PythonFastAPILangChainGPT-4o Vision

Client

Flutter WebFlutter Mobile

Data

Supabase (Postgres)pgvector

Infra

Firebase HostingCloudflare

My role

Solo project, end to end. Identified the problem from personal experience with the family business, designed the data model, built the FastAPI backend and LangChain pipeline, wrote the Flutter client, and deployed to Firebase. The app runs on real production data.