AgriFusion-ai

AgriFusion-AI is an AI-powered agricultural SaaS that helps farmers make smarter decisions from planting to market using photos, weather, soil data, local languages, and a trust-based marketplace.

Comment

Inspiration

Growing up in Nigeria, I watched my uncle lose nearly half his cassava harvest to a disease he couldn't identify. By the time he found an agricultural extension officer, it was too late. This story repeats across Africa — 250 million smallholder farmers struggle with crop diseases, limited veterinary access, and unpredictable markets. When I discovered Gemini 3's multimodal capabilities, I knew this was the tool that could finally bring expert agricultural knowledge directly to farmers' phones, in their own language.

What it does

AgriFusion AI is an all-in-one farm management platform powered by Gemini 3 Flash. Farmers can:

  • Diagnose plant diseases by snapping a photo — get instant treatment recommendations with urgency levels
  • Analyze soil health using Gemini vision combined with Open-Meteo satellite data for pH, moisture, and crop recommendations
  • Get AI veterinary advice for livestock with multi-turn conversations
  • Speak in Nigerian languages — Voice AI with Yoruba, Hausa, Igbo, and Pidgin support using authentic local voices
  • Track disease outbreaks on an interactive map with community and API-fed alerts
  • Trade in the marketplace — buy/sell crops, seeds, and livestock with in-app chat
  • Manage livestock with breeding guides, vaccination schedules, and feeding plans
  • Access microloans through Farm Score credit system with Paystack integration
  • Discover trends — trending crops, top farmers, and live market prices

How we built it

We built AgriFusion AI with Next.js 15 and TypeScript for the frontend, styled with a mobile-first approach. The backend uses Supabase for authentication and database, Firebase for push notifications, and Paystack for wallet and payment integration.

The AI features are powered by Gemini 3 Flash with a smart fallback chain to Gemini 2.5 and 2.0 for reliability. We integrated Open-Meteo satellite APIs for real-time weather and soil moisture data. Voice features use YarnGPT for Nigerian language transcription and text-to-speech.

My partner contributed the chatbot assistant foundation and the live market trend integration, while I built everything else — from the complete UI/UX to all 9+ Gemini-powered API endpoints.

Challenges we ran into

  • Making the demo video: This was the hardest task — editing screen recordings, syncing voiceover with slides, cutting unwanted segments, and combining everything into a cohesive 3-minute video required countless iterations.
  • Limited disease outbreak APIs: Finding reliable public APIs for agricultural disease data in Africa was extremely difficult.
  • No veterinarian location API: We couldn't find an API for veterinarians in Nigeria.
  • Nigerian voice AI: Getting speech-to-text working for Yoruba, Hausa, Igbo, and Pidgin was challenging.
  • Mobile-first with complex features: Balancing a rich feature set with mobile responsiveness.
  • Gemini response parsing: Ensuring consistent JSON output from Gemini required careful prompt engineering.

Accomplishments that we're proud of

  • Mobile-first responsive design that works beautifully on any screen size
  • Dark mode support throughout the entire application
  • AI Soil Analysis combining Gemini vision with satellite weather data — a unique feature we haven't seen elsewhere
  • P2P Marketplace with bidding, exchange, and in-app chat with image sharing
  • Voice AI in Nigerian languages with authentic local voices (Idera, Zainab, Femi, Adaora)
  • Disease outbreak map providing early warning to farmers

What we learned

  • Prompt engineering is critical: Getting consistent, structured JSON responses from Gemini required extensive testing and fallback handling
  • Multimodal AI is game-changing: Combining image analysis with contextual data (GPS, weather) creates powerful features not possible before
  • African language AI needs work: There's a huge gap in Nigerian language support that we had to work around
  • Mobile-first mindset: Designing for farmers in rural areas with limited data means every kilobyte and tap matters
  • Community-powered data fills gaps: When APIs don't exist, user-contributed data can be just as valuable

What's next for AgriFusion AI

  • Production launch: Acquiring a custom domain, proper hosting, and scaling infrastructure
  • Google Translate API integration: Auto-translate the entire UI to local languages
  • Microloan partnerships: Partner with agricultural microfinance institutions to offer loans directly through Farm Score
  • Offline mode: Allow key features to work without internet for remote areas
  • Expand to other African countries: Localize for Kenya, Ghana, and Tanzania markets

Built With

Share this project:

Updates