AgriAI: Turning A Farmer’s Phone into an AI Cocoa Advisor
For the Small AI for Development Hackathon organized by the World Bank Group and Hack-Nation, a group of CodeLink team members developed AgriAI, an Android cocoa farming assistant that combines offline disease detection, AI-powered guidance, and French voice interaction with SMS access to price and weather information.
2 Days
From research and design to a working AI prototype, following the Hackathon’s timeline.
8,950 Images
Public cocoa disease images used to train the computer vision model.
14,660 Annotations
Labeled bounding boxes across the training dataset.
7 Cocoa Conditions
Image classification categories covering healthy plants, diseases, and pest damage.
3 Offline AI Capabilities
Computer vision, knowledge-grounded AI assistance, and French speech interaction running directly on Android.
7 Knowledge Cards
Structured agricultural guidance used to ground AI-generated answers.
About the client
The Small AI for Development Hackathon challenged participants to develop practical, lightweight AI solutions addressing real-world development challenges. The event emphasized making AI accessible in resource-constrained environments, where internet connectivity, device performance, and digital literacy can limit technology adoption.
The challenge
Smallholder cocoa farmers in Côte d'Ivoire and Papua New Guinea face limited internet access, infrequent agricultural support, and significant crop disease risks. CodeLink set out to explore how lightweight, accessible AI could help farmers identify crop problems and make more informed decisions without relying on constant connectivity.
Why CodeLink?
CodeLink combined product research, user-centered design, on-device AI development, and full-stack engineering to tackle the challenge. With expertise spanning Android, computer vision, generative AI, and backend systems, our team transformed an agricultural use case into a working prototype within 48 hours.
The result
Completed during the Hackathon, AgriAI was submitted as a working prototype. It combines offline crop disease detection, knowledge-grounded AI guidance, and French voice interaction. An SMS hotline provides access to price and weather information without mobile data. Built in just two days, the solution demonstrates CodeLink's ability to integrate multiple on-device AI technologies into a functional, offline-first agricultural assistant designed for farmers with limited connectivity.
