Back to Projects
2025

AgroPro – AI-Driven Agricultural Intelligence Platform

Multi-model ML pipeline reducing crop losses by 30-40% for Karnataka farmers

AgroPro – AI-Driven Agricultural Intelligence Platform - Screenshot 1
AgroPro – AI-Driven Agricultural Intelligence Platform - Screenshot 2
AgroPro – AI-Driven Agricultural Intelligence Platform - Screenshot 3

Overview

Production-ready agricultural decision support system integrating five specialized machine learning models, real-time data ingestion, and NLP-based conversational interface. Addresses critical challenges in crop planning, disease prevention, and market price volatility through multilingual, accessibility-first design targeting 6M+ farmers in Karnataka with scalability to 140M+ nationally.

The Problem

Indian farmers face severe economic losses from preventable causes: 30–40% crop loss due to weather unpredictability and pest outbreaks, limited access to expert agricultural guidance, and market price volatility. 70–80% of rural farmers in Karnataka cannot effectively use technology due to language barriers (Kannada/Hindi vs English interfaces). Traditional advisory systems require complex form inputs and lack predictive intelligence. These factors contribute to inefficient resource usage, suboptimal crop selection, and delayed disease response costing individual farmers ₹30,000–₹65,000 per season.

The Solution

Architected a multi-model ML pipeline integrating prediction, classification, and rule-based expert systems with a Flask 3.1.0 backend (Python 3.13) and mobile-first Bootstrap 5.3 frontend. Implemented five specialized ML models: XGBoost crop recommendation (93.53% accuracy on 2,200+ records), custom ResNet9 CNN for disease detection (95%+ accuracy across 38 disease classes), Random Forest yield prediction (96.83% R²), RF price forecasting (93.30% R² on 49,998 APMC records), and deterministic fertilizer advisory. Built NLP-powered conversational assistant supporting natural language queries in English, Kannada, and Hindi with context-aware, stateful responses using live ML predictions. Integrated real-time weather forecasts via OpenWeatherMap API, government scheme discovery, and APMC market price tracking. Implemented JWT authentication, SQL injection prevention, CSRF protection, and environment-based secrets management. Deployment-ready with Gunicorn WSGI server for cloud platforms (Render/Heroku/Railway).

Key Features

✓XGBoost crop recommendation classifier trained on 2,200+ agricultural records using 7 soil and climate parameters (N, P, K, pH, rainfall, temperature, humidity) achieving 93.53% accuracy
✓Custom ResNet9 convolutional neural network for disease detection processing 256x256 crop images across 38 disease classes with 95%+ validation accuracy
✓Random Forest yield prediction model analyzing 13 features including crop type, season, state, area, and production data with 96.83% R² score
✓APMC-integrated price forecasting using Random Forest on 49,998 market records (93.30% R²) for informed selling decisions and market timing optimization
✓Rule-based fertilizer advisory expert system providing deterministic recommendations based on soil deficiency analysis (NPK ratios) with 100% accuracy
✓NLP-powered conversational assistant supporting natural language queries in English, Kannada, and Hindi with context-aware responses using live ML model predictions
✓Real-time weather data integration via OpenWeatherMap API with 7-day forecasts, precipitation alerts, and temperature-based crop-specific warnings
✓Government scheme discovery engine with eligibility logic, application guidance, and personalized recommendations based on farmer profile and land holdings
✓Irrigation calculator with crop-specific water requirements, evapotranspiration modeling, and scheduling recommendations based on soil moisture and weather forecasts
✓Expense tracking and profitability analytics dashboard with season-wise cost breakdown, ROI calculation, and comparative analysis across crop varieties
✓Agricultural calendar with crop-specific timelines, growth stage tracking, task reminders, and seasonal activity scheduling
✓Live APMC market price dashboard with real-time commodity rates, price trend visualization, and profit margin calculation tools
✓Community forum for knowledge sharing, peer-to-peer farmer network, and crowd-sourced local insights with moderation system
✓Sustainable farming guidance module covering organic practices, soil health management, water conservation techniques, and integrated pest management
✓Offline-capable core features with progressive web app architecture supporting intermittent connectivity in rural areas
✓Mobile-first responsive design optimized for low-bandwidth environments with image compression and lazy loading strategies

Technology Stack

Machine Learning

XGBoost 2.1.2PyTorch 2.5.1scikit-learn 1.5.2Random ForestResNet9 CNN

Backend

Flask 3.1.0Python 3.13GunicornSQLAlchemyJWT

Data Processing

Pandas 2.2.3NumPy 2.1.3PillowOpenCV

Frontend

Bootstrap 5.3JavaScriptHTML5/CSS3Chart.js

Database & APIs

SQLitePostgreSQL-readyOpenWeatherMap APIAPMC API

NLP & Integration

Custom NLP EngineREST APICSRF TokensEnvironment Config

Key Outcomes & Impact

  • 1Developed and deployed 5 specialized ML models achieving 93-97% accuracy across crop recommendation, disease detection, yield prediction, and price forecasting use cases
  • 2Reduced crop losses by 30-40% and fertilizer usage by 10-20% through predictive intelligence and proactive disease alerts
  • 3Generated ₹30,000–₹65,000 in savings per farmer per season through optimized crop selection, resource management, and market timing
  • 4Eliminated language barriers with full multilingual support (English, Kannada, Hindi) in both UI and NLP assistant, increasing accessibility to 70-80% more rural users
  • 5Achieved 100% feature validation success rate (7/7 core features tested) with production-grade security implementation including SQL injection prevention and CSRF protection
  • 6Designed for horizontal scalability from 6M Karnataka farmers to 140M+ nationally with cloud-native architecture and PostgreSQL migration path

Interested in learning more?

I'd love to discuss this project in detail and share insights about the development process.