Applied ai
Krishyak
FastAPI
React
TensorFlow
Problem
The problem
Farm decisions bring together uncertain yields, market prices, weather, and crop health. Farmers need to compare choices in a language they can use.
Approach
How I approached it
I built a full-stack decision platform combining yield and risk estimation, price forecasting, government-scheme matching, fertilizer guidance, and voice input. Monte Carlo simulations let users compare 100–2,000 possible scenarios. The interface supports 13 Indian languages.
Details
What I built
- Trained a MobileNetV2 plant-disease classifier across 42 classes using approximately 70,000 images.
- Recorded 95.2% test accuracy and F1 0.94 in the supplied resume; these are model test results, not a guarantee for field diagnoses.
- Integrated crop-aware treatment and prevention guidance.
- Presented Krishyak at VBYLD 2026 to the Union Agriculture Minister and the Director General of ICAR.