mikeledusei/AgriShield

Protecting Kenya's food security, ensuring future sustainability. AgriShield is a hierarchical Agentic AI system that predicts crop yield shocks and livestock forage deficits across all 47 counties turning XGBoost forecasts into plain-English insights, live maps, and automated PDF reports for farmers and government. Protecting Kenya's Food Security, Ensuring Future Sustainability. **A Hierarchical, Agentic AI Predictive Intelligence System for Crop Yield and Livestock Forage Risk in Kenya** --- ## 📖 Table of Contents - About the Project - The Problem - The Solution - Key Features - System Architecture - Tech Stack - Project Structure - Getting Started - Data Sources - The Team - The 5 Critical Questions - Roadmap - Documentation - License - Acknowledgments --- ## 🎯 About the Project **AgriShield** is a predictive intelligence system that forecasts agricultural risks across Kenyan counties before disasters happen. It predicts both **crop yield shocks** and **livestock forage deficits**, then uses an AI assistant named **Gria** to translate complex predictions into plain-English insights, dynamic maps, and professional PDF reports. The system works at two levels: - **County Level:** Detailed risk predictions for individual counties. - **Regional Level:** Aggregated intelligence for entire regions like the Rift Valley or Eastern Kenya. > AgriShield shifts agricultural disaster management from **reactive** (responding after crops fail) to **proactive** (acting before the damage occurs). --- ## ⚠️ The Problem Agriculture contributes about **33% of Kenya's GDP** and employs over **75% of the rural population**. Yet the sector faces serious challenges: 1. **Reactive Responses:** Early Warning Systems only activate *after* crops have already failed or livestock are already starving. 2. **Siloed Solutions:** Existing tools focus only on crops *or* only on livestock, never both together. 3. **Complex Outputs:** Current systems produce technical GIS maps that county officers cannot easily understand or act on. 4. **The "Black Box" Problem:** Machine learning models give raw numbers without explaining *why*, making it hard for officials to justify budgets and actions. The result is billions of shillings lost every year to preventable agricultural disasters. --- ## …