KYRA: Point a Camera at a Farmer and Know Everything

The Problem

Challenge

Here is the real problem: a farmer walks into an agri-input store. The field agent needs to know who this person is, what land they own, what crops they are growing, what they bought last season, and what fertilizer they might need. Doing this manually takes 5-10 minutes. During peak season, with 50+ farmers walking in daily, it is impossible. Sales get lost. Fertilizer allocation becomes guesswork.

The Approach

How I thought about it

AWS Rekognition handles the face matching. But the real work was connecting recognition to business data. When KYRA identifies a farmer, it queries the customer profile, land records, crop cycles, soil tests, and purchase history all at once. It compiles everything into a single view. The agent sees everything they need to make a sale. The system calculates estimated fertilizer demand based on land area, current crops, and soil conditions. No more guessing.

The Build

What was built

• Built face enrollment flow: capture photo → link to CRM profile → index in Rekognition. Takes 30 seconds per farmer.
• On recognition, queries 6+ database tables simultaneously and surfaces a complete profile: who they are, their land, crops, soil, irrigation, purchases, pending orders.
• Designed the UI for speed: camera capture button, fallback manual search, everything on one screen.
• Fertilizer demand estimation based on actual crop data and soil health. Not guesswork.
• Offline queue for areas with bad connectivity. Photos queue locally, sync when internet returns.
• WhatsApp integration: recognition triggers automated personalized messages and discount cards.

The Outcome

Results

Tech Stack

Technologies used