← Junaid Afzal All work
In progress · expected Feb 2027 Final Year Project

Farm2Fork

A farmer-to-buyer marketplace with blockchain provenance and AI crop grading

View on GitHub ↗

5
Repositories: web, backend, mobile, blockchain, AI
4
Crops graded: wheat, rice, mango, maize
A–D
Quality grade scale
Role
Lead developer across web, backend, mobile, blockchain and AI
Company
University of Central Punjab
Period
Apr 2026 – Feb 2027

01

Overview

Farm2Fork connects farmers directly with buyers, with transporters handling delivery. It is organised as five repositories in a GitHub organisation — web, backend, mobile, blockchain and AI — and I am the lead developer across all of them.

02

System map

  1. 01

    Web

    Next.js · TypeScript

  2. 02

    Backend

    NestJS · MongoDB · Redis

  3. 03

    Mobile

    Flutter · Google Maps · FCM

  4. 04

    Blockchain

    Hyperledger Fabric · Go chaincode

  5. 05

    AI

    FastAPI · EfficientNet-B0

03

How it is built

  1. 01

    Backend

    A NestJS API on MongoDB and Redis with Firebase authentication, Swagger documentation, Docker Compose to run the whole stack locally, and end-to-end tests covering the payment and shipment flows.

  2. 02

    Web

    A Next.js app with separate routes for buyers, farmers and transporters, Firebase sign-in (including Google) and credentialed, HTTP-only backend sessions protected by a CSRF header.

  3. 03

    Mobile

    A Flutter app with Google Maps for farm pins, the checkout drop-off pin and transporter maps; transporters share location only while online with the app open, and Firebase Cloud Messaging delivers push notifications.

  4. 04

    Blockchain

    A Hyperledger Fabric network with Go chaincode that records payments (Stripe, JazzCash, Safepay) and supply-chain events such as listed, assigned, picked up, in transit and delivered. Writes are validated and idempotent, and the ledger can be queried by ledger key, reference or product.

  5. 05

    AI

    A FastAPI inference service that suggests an A–D quality grade from a photo of wheat, rice, mango or maize (EfficientNet-B0) for the farmer to confirm, plus a rule-based price-range estimate in PKR per unit.

04

Engineering notes

The AI grade is a suggestion the farmer confirms, not a certified assessment, and the price estimate is rule-based until a trained model is available — the documentation says so explicitly, and so does this page. A benchmark script measures ledger write throughput at different concurrency levels.

05

Beyond the code

The project also includes market research, competitor analysis and SWOT analysis to design a go-to-market strategy, with proposed pricing, positioning and promotional plans aimed at market penetration and customer acquisition.

06

Stack

  • NestJS
  • MongoDB
  • Redis
  • Next.js
  • TypeScript
  • Flutter
  • Firebase Auth
  • Hyperledger Fabric
  • Go chaincode
  • FastAPI
  • Docker