Inspiration

Pakistan has 22 million wheat farmers, most with no access to agronomists. A single crop disease can destroy an entire season. We built Wheatee to put an AI agronomist in every farmer's pocket — speaking their language, Urdu.

What it does

Wheatee is an AI-powered farming assistant that:

  • Diagnoses crop diseases from photos using Gemini Vision AI
  • Answers farming questions in Urdu and English via voice or text
  • Remembers farm history, profiles, and past diagnoses
  • Provides fertilizer, irrigation, and yield advice personalized to each farm

How we built it

  • Mobile App: React Native + Expo for cross-platform iOS/Android
  • AI Agent: Google Agent Studio with Gemini 2.5 Pro, deployed via Vertex AI
  • Backend: Node.js + Express deployed on Google Cloud Run
  • Database: MongoDB Atlas for farm profiles, diagnoses, and conversations
  • MCP: Custom MCP server exposing 5 tools to the AI agent
  • Voice: Google Cloud Speech-to-Text for Urdu voice questions

Challenges we faced

  • Configuring MCP with Google Agent Studio (OAuth support was "coming soon")
  • Getting Gemini to consistently respond in Urdu
  • Deploying React Native with native modules (Mapbox) on Expo Go
  • Managing GCP IAM permissions across a 4-person remote team

What we learned

  • Google Agent Studio + MCP is powerful but still maturing
  • MongoDB Atlas MCP integration works great for agent memory
  • Urdu NLP with Gemini 2.5 Pro works surprisingly well
  • Cloud Run is incredibly fast to deploy and scale

What's next

  • Add satellite field mapping with real Mapbox integration
  • Weather API integration for planting recommendations
  • Offline mode for farmers with poor connectivity
  • Expand to other crops: rice, sugarcane, cotton

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Wheatee

Built With

Share this project:

Updates