THE PROBLEM: Millions of small and mid-scale farmers face financial losses caused by crop diseases, livestock illness, drought, floods, and market disruptions. While governments and NGOs offer support programs — grants, subsidies, emergency relief, and technical assistance — many farmers are unaware these opportunities exist or struggle to understand eligibility requirements and application processes. As a result, farmers miss out on critical resources that could help them recover, protect their livelihoods, and improve food security. This is not a lack of will — it is a lack of access and clarity.

OUR SOLUTION:
FarmGuard AI is an AI-powered support navigator that helps farmers in Africa understand which agricultural support programs they may qualify for and guides them through the next steps. A farmer creates a simple profile describing their farm, crops, animals, and current challenge. From there, FarmGuard AI personalizes everything: it asks relevant questions to interpret their situation, translates complex eligibility rules into plain language, and connects them to real government and NGO programs they may be eligible for. Crop and livestock disease diagnosis is built directly into this journey — when a farmer identifies a health problem on their farm, FarmGuard AI automatically surfaces support programs that may help cover treatment costs or connect them to emergency relief, turning a moment of crisis into a clear path forward.

HOW IT WORKS:
The farmer provides input through a personalized profile and intake — farm size, years of farming, crops and animals raised, current challenges, and urgency. Google Gemini Vision AI analyzes photos of crops or animals when diagnosis is needed, identifying multiple issues per photo with severity and confidence ratings, and explaining its reasoning in plain language. For support navigation, the AI interprets the farmer's full situation against known program criteria, generates relevant eligibility questions specific to each program type, and gives a cautious "may be eligible" assessment — never a guarantee. Every interaction is personalized using the farmer's saved profile, so they are never asked to repeat themselves.

WHO WE BUILT THIS FOR:
Small and mid-scale farmers across Zambia and developing regions in Africa — people operating under time pressure and financial stress, with limited literacy, limited internet access, and no affordable way to reach professional agricultural or government support when they need it most.

WHAT MAKES FARMGUARD AI DIFFERENT: FarmGuard AI does not just display a list of programs — it interprets each farmer's specific situation and guides them to a clear next step. It explains its own reasoning for every diagnosis and every program match. It generates realistic, properly paced treatment plans matching real farming urgency, not token daily tasks. It connects farmers to a real WhatsApp farming community, local expert categories, and lets them download a professional PDF report of their diagnosis to share with officials or experts. A 4-digit PIN with secure recovery protects each farmer's personal data, since this information is sensitive and private.

RESPONSIBLE AI: Every result uses cautious, non-guaranteed language — "this appears to be," "you may be eligible" — never definitive claims. Every diagnosis includes a confidence level and a visible disclaimer that results are guidance only. High severity cases are flagged urgently with direct expert contact options via WhatsApp, SMS, and email. The AI never makes the final decision — the farmer does, supported by a qualified human expert for serious cases. Farmer photos and data are never sent to or stored on any external server; everything stays securely on the farmer's own device, protected by their PIN.

IMPACT: FarmGuard AI reduces friction in navigating two interconnected systems farmers depend on: agricultural health knowledge and public support programs. It helps farmers move from confusion to clarity to action — protecting harvests, saving animals, and improving access to assistance that already exists but was previously too hard to find.

Built With

Share this project:

Updates