Cyrax
Khetsmart - Agri-Fintech
Submitted
May 22, 2026, 6:56 AM
Last Updated
May 22, 2026, 6:56 AM
Project Links
Problem Statement
West Bengal’s potato farmers frequently face distress selling due to limited cold storage capacity, inefficient logistics, lack of market intelligence, and delayed financial support. During bumper harvest seasons, large quantities of potatoes enter the market simultaneously, causing severe price crashes. As a result, farmers are often forced to sell crops below cultivation cost to avoid spoilage and repay debts. The existing agricultural supply chain is reactive and poorly coordinated, leading to storage overload, inefficient distribution, financial losses, and unstable farmer incomes. This recurring “Harvest Trap” turns successful harvests into economic crises for farming communities.
Solution & AI Usage
KhetSmart is an AI-powered Agri-FinTech platform designed to predict agricultural oversupply and intelligently coordinate storage, logistics, and financing before market prices collapse. The system was built using satellite intelligence, machine learning, and graph-based optimization models. For crop prediction, we used Sentinel-2 satellite imagery and NDVI vegetation analysis with computer vision models built using TensorFlow and Python to forecast potato yield and detect possible market gluts early. For dynamic storage allocation and transport optimization, we implemented Graph Neural Networks (GNNs) using PyTorch Geometric to analyze cold storage availability, farmer location, and logistics routes in real time. The platform also integrates embedded financing through digital warehouse receipt systems, enabling farmers to access short-term liquidity without distress selling. APIs such as Google Maps API and weather datasets were used for routing and environmental analysis.
Full Description
KhetSmart — Agri-FinTech for West Bengal Potato Farmers
Problem Every harvest season, potato growers in West Bengal face the same chain of losses: oversupply drives mandi prices down, farmers sell under pressure at distress rates, and cold-storage capacity exists nearby but is hard to find and book. Transport and micro-loan costs are unclear, so “profit” is guessed from agent talk or WhatsApp forwards—not from live mandi data, satellite signals, or a route to the right godown. Information is fragmented, not available in Bengali/Hindi at the field, and not tied to the farmer’s GPS, quantity, or potato variety.
Solution KhetSmart is a mobile-first Progressive Web App that turns real agricultural and market data into one harvest plan: signal → route → cold storage → payment → optional mandi auction. Farmers see whether the season is risky (glut), what they may earn at live corridor mandi prices (adjusted by crop type and quintals), the best cold-storage path from their farm location, and a micro-loan offer when market pressure is high. The platform is built for low-literacy use: large chips, voice/text input, and Bengali, Hindi, and English UI.
Farmer tab Users quick-pick harvest size (25–100 quintals) and potato type (e.g. Jyoti, Chipsona). Profit and live ₹/quintal update automatically when selections change, using ingested mandi data blended with crop-specific factors—not a single hardcoded figure. “Get route + loan” runs an optimizer over hundreds of storages (OSRM/haversine routing), shows glut insight, transport cost, and loan eligibility. Farmers book cold storage with Razorpay (UPI, card, or COD), receive a downloadable receipt, and—when logged in via OTP—save orders under My Orders with total quintals tracked on profile.
Predict tab A “Potato market watch” shows glut risk, mandi price signals, and storage fill pressure. A live weather panel pulls OpenWeather (current conditions, humidity, wind, 5-day forecast) merged with Open-Meteo corridor statistics. Gemini AI Crop Brain generates advisory text from live vegetation indices (NDVI, SAVI, GNDVI), weather, soil, and mandi layers—not static placeholder copy when API keys are configured.
Network tab An interactive map of 496+ cold storages across the potato corridor shows utilization, critical sites, spare capacity, and district filters—reducing blind travel during harvest peaks.
Finance tab Loan and insurance pathways support liquidity and risk cover. The Auction module lets farmers list only their own confirmed cold-storage bookings for mandi-style bidding; they can modify start price or withdraw listings. Farmers do not bid on others’ lots—buyers compete on the farmer’s harvest.
Technology Frontend: React, TypeScript, PWA. Backend: FastAPI, SQLAlchemy (SQLite/Postgres), farmer OTP auth, booking and auction services, Razorpay integration. Data: data.gov.in mandi ingest, OpenWeather, Open-Meteo, Copernicus/Sentinel-style NDVI pipeline, ERA5 soil moisture, OSRM routing, optional Google
Tech Stack
Screenshots
4 attached
