AgriVision AI — Crop Disease Detection Platform
End-to-end agricultural AI platform where farmers upload crop images and receive ML-powered disease diagnoses in under 60 seconds
Project Overview
AgriVision AI is a full-stack agricultural decision-support system built to reduce crop disease diagnosis time from days to under 60 seconds. Farmers upload crop images through a React frontend, which triggers a Node.js (NestJS) backend to orchestrate ML inference via a Python FastAPI service running an EfficientNet model trained on 38+ disease classes. Diagnosis results include confidence scores and expert verification flows. The backend features 8 domain-separated service modules, JWT authentication with RBAC for farmer and expert roles, email verification, password reset, and composite-indexed PostgreSQL schemas optimized for scan history, disease records, and reporting.
Core Challenges
- [1]Orchestrating HTTP multipart calls from Node.js to FastAPI ML inference service reliably under varying image sizes
- [2]Designing role-based access control (farmer/expert) with email verification and password reset flows from scratch
- [3]Building composite-indexed PostgreSQL schemas for disease records, scan history, and reports to keep query performance fast at scale
Key Outcomes & Outcomes
- ✓Reduced crop disease diagnosis turnaround from several days to under 60 seconds
- ✓Achieved zero authentication-related vulnerabilities across automated security scans
- ✓Cut p95 API response time by 35% through parameterized queries and composite indexes on scan and prediction tables
- ✓Integrated ML inference service covering 38+ disease classes via EfficientNet/timm model