Back to projects
aiProductionPRODUCTION

AgriSahayak

AI-Powered Crop Disease Detection Agent

WhatsApp-based farming assistant using Gemini Vision AI for real-time crop disease detection, treatment plans, and weather alerts. Bilingual Urdu/English.

GitHub
Case study snapshot
Role
AI/product builder
Audience
Pakistani farmers, hackathon judges, and agri-tech stakeholders
Timeline
2025-09
Role
AI/product builder
Audience
Pakistani farmers, hackathon judges, and agri-tech stakeholders
Date
2025-09
Problem

Why it mattered

Small farmers need fast crop disease diagnosis, localized treatment, and supplier guidance, but expert help can take days to reach them.

Solution

What shipped

A multi-agent farming assistant that accepts crop images, diagnoses likely disease, explains treatment in simple language, and connects the farmer to nearby support.

Impact

Proof signal

60% projected crop loss reduction

< architecture />

System flow

Image upload enters a vision agent, the diagnosis is passed into a treatment planner, weather and supplier context are layered in, and the response is returned through a farmer-friendly interface.

Vision Input
Disease Classifier
Treatment Planner
Alert System
Visual architecture
01
Vision Input
02
Disease Classifier
03
Treatment Planner
04
Alert System
quality score

Honest maturity snapshot

Code Quality8/10
UI / UX6/10
Scalability7/10
Production Ready8/10
Before

Manual questions, scattered context, and slow follow-up.

After

An agent workflow that routes intent, calls tools, and returns useful next actions.

capabilities

Features

  • Gemini Vision diagnosis for crop leaf images.
  • Localized treatment plan with practical next actions.
  • Bilingual Urdu/English communication flow.
  • Supplier discovery designed around a local radius.
  • Weather-aware recommendation layer.
  • Hackathon-ready demo flow with clear farmer value.
tradeoffs

Engineering challenges

  • Balancing confident diagnosis with safe uncertainty language.
  • Keeping the UX simple for users who may not be technical.
  • Turning a broad agricultural problem into a focused hackathon MVP.
proof

Results

  • Won Innovista Agentic AI Hackathon.
  • Projected crop loss reduction story made the business value easy to understand.
  • Became the strongest flagship case study for agentic AI plus social impact.
reflection

What I learned

  • Tradeoff learned: Balancing confident diagnosis with safe uncertainty language.
  • Proof learned: Won Innovista Agentic AI Hackathon.
  • Next iteration: Add agronomist review workflow for high-risk diagnoses.
< stack />
Gemini Visionn8nWhatsApp APIPython
< next iteration />

Future improvements

Add agronomist review workflow for high-risk diagnoses.
Store anonymized crop cases for regional disease trend analysis.
Add offline-first guidance and voice notes for low-literacy users.