University project · Plant health
AgriAI
An AI plant-health and diagnostic system — an overhead camera or uploaded photos feed a Gemini model that scores health, flags disease and nutrient deficiency, and rolls it up into per-section dashboards.
- Role
- Product & delivery, solo
- Build
- AI-assisted · Gemini API
- Type
- University project
- Status
- Live demo
The idea
Growers walk their fields and catch problems late. AgriAI is a take on doing it continuously: an overhead camera — or a phone photo — feeds an AI model that scores each plant's health, flags nutrient deficiencies and disease, and rolls the results up into a dashboard per crop section.
How I approached it
I designed the whole loop: image ingestion (camera or upload), the diagnostic call to Google's Gemini model, health scoring, a fertilizer / nutrient status breakdown (Balanced, N-Deficient, Chlorosis, K-Burn), a per-section CMS, and a simulation engine to generate test data — all built with AI to my direction. The product questions were about actionability — what does a grower actually do with a “72% health score,” and when is a flag worth walking out to check?
What's there
A working demo at the live app: health metrics and diagnostic history across crop sections, a health-score distribution, fertilizer / nutrient status, a visual log grid of scanned plants, plus image ingestion and a section manager. Diagnostics run through Gemini.
What I'd do next
Ground the model on a labeled dataset for one crop and one region, and measure diagnosis accuracy against an agronomist before trusting any of the numbers on the dashboard.