Inspiration
During a semester break, my friends and I — all computer science students — returned to our rural hometown in Ethiopia, one of the country's top wheat-producing areas. We were shocked to see widespread crop loss due to diseases that could have been treated early.
Local agro-extension workers were using AI-based applications to diagnose plant diseases, but these tools were trained on non-local, global datasets. As a result, they misdiagnosed local conditions, recommended the wrong treatments, and actually made things worse.
This personal experience inspired us to build a locally trained AI system — one designed to understand Ethiopian crop diseases, operate offline, and provide real-world, field-tested solutions to farmers like our own families.
What it does
Dr. Azmeraw is an AI-powered crop health management system designed for low-resource settings. It:
Diagnoses plant diseases using locally trained deep learning models Works through a mobile app and Telegram bot — even offline Provides tailored, context-aware treatment recommendations Tracks infected areas via GPS for precision spraying Supports local languages and works on low-end smartphones
It empowers farmers to detect early, act fast, and save crops — anywhere, anytime.
How we built it
We followed a highly localized, iterative development approach: Collected local wheat disease images and expert diagnoses Built a custom AI model trained specifically on Ethiopian conditions Developed a mobile app and Telegram bot for field usage Integrated offline functionality and support for low-end devices Collaborated with Jimma University and local agri-experts Piloted and refined the system through hands-on feedback from real farmers
Challenges we ran into
Data Scarcity: There were no quality datasets for Ethiopian wheat diseases — we had to collect and annotate our own. Farmer Skepticism: Previous failures of global tools meant low trust. We overcame this by co-designing with local farmers. Low Infrastructure: Rural areas lack stable power, fast internet, or modern devices — we built with these constraints in mind. Funding Gaps: We bootstrapped much of the early work and are still seeking sustainable funding to scale.
Accomplishments that we're proud of
✅ Developed and field-tested an AI model capable of accurately detecting common Ethiopian crop diseases ✅ Built a mobile app + Telegram bot that works offline and in local languages ✅ Partnered with farmer cooperatives, agro-dealers, and agricultural offices ✅ Secured interest from researchers and policymakers ✅ Initiated real-world impact, with positive feedback from early adopters
What we learned
- Localization is key: AI must be retrained and re-evaluated for every region
- Offline-first is a necessity: For rural Africa, connectivity cannot be assumed
- Community collaboration matters: Trust is built through transparency, listening, and iteration
- AI isn't just code — it’s culture, context, and communication
What's next for Dr. Azmeraw: AI-Based Crop Health Management System
🎯 Scale the system to cover other major crops (maize, teff, sorghum) 🤝 Deepen partnerships with agricultural institutions and ministries 🚁 Integrate with drone spraying services for precise, GPS-based interventions 🌍 Expand across Ethiopia, then into East Africa and other low-resource regions 📈 Empower millions of farmers with affordable, accessible AI that saves food and improves livelihoods
Log in or sign up for Devpost to join the conversation.