Portfolio

~/projects/khetibaadi.mdsource ↗

Khetibaadi

#flask #react #node-js #postgresql #docker #keras #pytorch #opencv

A GIS-based farm planning platform built to help farmers make data-driven decisions — crop recommendation, yield prediction, fertilizer guidance, and plant disease detection, wrapped in a multilingual web app with an NLP chatbot for onboarding.

Architecture

A Flask-based ML microservice sits behind the main Node.js/React app, exposing a RESTful API that the frontend calls for crop prediction, geo-location lookups, and chatbot responses — keeping the model-serving layer independent of the main application.

Models

  • Crop yield prediction
  • Fertilizer recommendation
  • Plant disease detection — a ResNet CNN reaching 98% accuracy
  • NLP-powered chatbot for guiding users through the app

Infrastructure

  • PostgreSQL for data persistence
  • Docker containers for backend deployment

Stack

  • Frontend: React
  • Backend: Node.js, Flask (ML microservice)
  • ML/DL: Keras, PyTorch, OpenCV
  • Data: PostgreSQL
  • Infra: Docker

Source code