eco AgriSync

The field finally speaks data.

A precision farming system built in Uttar Pradesh. Free satellite and weather data on a single Raspberry Pi — warning farmers of disease and crop stress 3 to 7 days early, straight to their WhatsApp.

Dashboard arrow_forward

Explore

The AgriSync v3.0 Pipeline

Satellite-dominant, hardware-light. One Raspberry Pi runs the entire daily pipeline at the edge.

satellite_alt 01 / Satellites

Free Public Imagery

Sentinel-2 optical images every 5 days track crop health. Sentinel-1 radar sees through clouds to read soil moisture. Both free, no field hardware needed.

cloud 02 / Weather

NASA POWER + Open-Meteo

Daily temperature, rainfall and humidity from NASA POWER. Forecasts from Open-Meteo's ECMWF AIFS model. Both free, pulled each morning at 00:30 IST.

developer_board 03 / Edge Brain

One Raspberry Pi 5

One Raspberry Pi 5 in the village does everything: pulls data, runs models, sends messages. No cloud, no laptop, no office.

schedule 04 / Daily Pipeline

Eleven Scheduled Jobs

Each night the Pi fetches weather and satellite data, runs four models, and sends advisories by 06:00 IST. Scheduled jobs handle the full sequence.

database 05 / Local Storage

SQLite + DuckDB

Daily records go into a local SQLite database on the Pi. Heavier queries run on DuckDB. Nothing leaves the device by default.

chat 06 / Output

WhatsApp + Photos Back

Farmers get a short Hindi advisory on WhatsApp at 06:00 IST via Twilio. They can reply with a photo; a vision model on the Pi diagnoses it. Validated photos earn a small UPI payout via Razorpay.

Prediction v3.0

The PhytoPlex
Intelligence Stack

Four models, each doing one job well. Soil moisture from satellites and weather, crop stage from growing-degree days, disease risk from a three-method ensemble, and on-device photo diagnosis when the farmer sends a picture.

72h

Soil moisture horizon

3–7d

Disease lead time

Models in the stack

Pi 5 runs it all

neurology

Four Models Running On The Pi

A · B · C · D · v3.0 PhytoPlex

Model A · Soil Moisture

XGBoost Regressor

Forecasts soil wetness now and for the next 72 hours, using Sentinel-1 radar, Sentinel-2 moisture data, rainfall, and crop water loss from NASA POWER.

Model B · Crop Stage

Growing-Degree Days

Tracks cabbage from transplant to harvest by adding up daily heat. Simple math, no AI, accurate within a few days.

Model C · Disease Risk

Rules + Matched Filter + BOCPD

Three signals catch disease early: weather-based rules, a pattern-match on satellite images, and an alert when the field starts behaving oddly.

Model D · Photo Diagnosis

Moondream 2 Vision Model

A small open vision model runs on the Pi. The farmer sends a WhatsApp photo, Moondream 2 names the symptom and suggests treatment in Hindi within 90 seconds.

Crop Intelligence

Built for Banda Gobi

Cabbage-specific disease intelligence tuned to the Indo-Gangetic Plain

PhytoPlex tracks the four diseases that actually cost a Harpur cabbage farmer money. Each one has its own weather rule, its own satellite-index pattern, and its own advisory in Hindi.

Alternaria brassicicola

Alternaria Leaf Spot

Daytime 20–30°C with humidity above 90% and 9+ hours of leaf wetness. Hits late vegetative stage hardest. 3–7 day lead.

Hyaloperonospora parasitica

Downy Mildew

Cool 8–16°C nights, humidity above 85%, leaf wet for 4+ hours, and morning fog. Peak risk October–November.

Xanthomonas campestris

Black Rot

Warm, humid weather above 25°C with rain splash spreading the bacteria. Hits hardest around head formation.

Pectobacterium / Erwinia

Soft Rot

Waterlogged soil during head fill and storage. The advisory recommends drainage and harvest timing, not a spray.

The Resolve

Built from a village,
designed for millions.

cottage

Hyper-local Roots

Built for a real cabbage field in Harpur, eastern UP. Every weather rule, disease window and Hindi message is tuned for that farmer and his neighbours, not a generic crop.

language

Scales By Cluster

One Pi can serve 20 to 50 fields in a village cluster. The next village just needs another Pi — no new hardware in any farmer's field.

handshake

Free Data, Open Tools

Satellite imagery, weather and the vision model are all free or open source. The farmer pays nothing and even earns a small UPI reward for sending useful photos.

developer_mode

Integrated Logic

Village Field #01

Built on: iPad + Raspberry Pi 5. No laptop.

Field intelligence,

redefined.

The whole pipeline on one Pi.

Coded on single machine, no cloud bills, no field hardware to maintain — just the satellites overhead and one device on the wall.

A solo precision agriculture project from a village in India. Proving that real problems can be solved with free public data and whatever device is already on the desk.

Under the Hood

The AgriSync Tech Stack

developer_board Raspberry Pi 5

terminal Python 3.11

schedule systemd

database SQLite

storage DuckDB

api Flask

model_training XGBoost

visibility Moondream 2

satellite_alt Sentinel-1 / 2

wb_sunny NASA POWER

cloud Open-Meteo

chat Twilio WhatsApp

payments Razorpay UPI

Want to follow the build?

PhytoPlex is being built in the open for one cabbage field this season. Get in touch if you grow Brassicas, work on edge ML, or want to bring this to another village.

CONNECT arrow_forward

API Reference Hardware Specs Deployment Guide