Multilingual Crop Advisory for Smallholder Farmers
State Agricultural Extension Program
The Challenge
Agricultural extension officers in rural Maharashtra were struggling to reach 50,000+ farmers with timely crop advisories. Existing systems required farmers to navigate complex apps in English or Hindi, while most farmers in the region spoke Marathi and had limited smartphone literacy. Critical pest outbreak warnings were reaching farmers too late, resulting in preventable crop losses.
Our Approach
Voice-First Interface in Marathi
Built a WhatsApp-based voice bot that farmers could interact with in spoken Marathi. No app downloads, no typing required — just speak naturally.
Hyperlocal Weather + Satellite Integration
Integrated IMD weather data with satellite imagery to generate block-level advisories. Farmers received alerts specific to their taluka, not generic state-level warnings.
RAG System for Agricultural Knowledge
Built a retrieval-augmented generation system over 10 years of agricultural research papers, KVK advisories, and local best practices. Every recommendation cites its source.
Human-in-the-Loop Validation
All AI-generated advisories reviewed by agricultural officers before broadcast. System learns from corrections to improve accuracy.
Results
12,000+
Farmers reached in pilot
3x
Faster advisory delivery
94%
Voice recognition accuracy in Marathi
40%
Reduction in pest-related crop loss (pilot villages)
"For the first time, farmers are getting information they can actually use, in a language they understand, before it's too late."
District Agricultural Officer
Pune Division