Case Study AgTech · IoT

Same-day farm data, even when the internet drops.

Coffee farms in the interior of Bahia collected field and weather data on paper. Paulo Cardoso worked on the IoT platform that replaced that loop: sensor towers relayed readings across a local mesh, a gateway synchronized them to Node.js APIs on AWS, and offline-first merging kept late data clean. Decisions that used to wait days started happening the same day.

Client
DrFarm
Sector
Agriculture · IoT
Engagement
Ingestion APIs & offline sync
Deployment
3 coffee farms
Platform
Sensor mesh · AWS
A coffee farm manager holding an orange field notebook beside sensor equipment
DrFarm: Offline-first IoT brought same-day field data to coffee farms with unreliable internet.
By the numbers
3 farmsRunning the platform
10 towersPer farm
5–7Sensors per tower
Days → same dayOwner data latency

The reading arrived after the decision had already been made.

Field and weather readings traveled by hand. By the time the owner saw them, the information was days old.

Irrigation and pest treatment were being decided against last week's reality. Some sectors were watered far more than they needed while others got less, because managers could not compare today's rain, soil and sector readings when the decision was due.

Connectivity was not a temporary inconvenience. Most sensor locations had no direct internet connection at all, so the platform had to work through it by design.

Towers relayed data locally, then one gateway carried it to the cloud.

Each farm had one sensor tower per hectare. A tower carried five to seven sensors and read roughly five times an hour. The platform tracked evapotranspiration, temperature, rainfall, CO2, humidity, shallow and deep soil readings, radiation and wind speed.

The towers formed a mesh network and relayed readings to the one point on the farm with internet access. That gateway synchronized the stream to Node.js ingestion APIs on AWS.

What kept the data clean

  • Devices kept collecting and storing readings locally when the network dropped
  • The app synchronized with the local gateway when connectivity returned
  • Duplicate readings were removed by timestamp
  • Late-arriving data merged without overwriting or double-counting the stream
Diagram showing offline DrFarm sensor towers relaying readings over a mesh to a farm gateway and then to Node.js ingestion APIs on AWS
Offline-first from the sensor tower to the cloud: collect locally, relay over the mesh, merge by timestamp.
A solar-powered DrFarm sensor tower deployed between coffee rows in Bahia
In the field

One tower covered each hectare.

The hardware kept recording beyond the reach of ordinary connectivity. A mesh carried the readings across the farm, and one gateway with cellular access synchronized them to the cloud.

10 towers per farm · 5 to 7 sensors each
Manager surfaces
DrFarm product design showing farm sectors drawn on a map and associated with sensor capsules
DrFarm product design showing threshold alerts by metric and farm sector

Product design screens from the project. Paulo's ingestion and offline-sync backend fed these surfaces.

Design render of a DrFarm sensor tower with solar panel, anemometer, rain gauge and capsule enclosure
The DrFarm gateway board with XBee mesh radio, cellular modem and Intel Edison

The field tower and the gateway board that centralized the mesh before cloud synchronization.

Engineering note

Offline-first meant treating disconnection as a normal state. Every reading needed an identity and a timestamp, so data arriving late could merge into the cloud stream without replacing a newer value or counting the same observation twice.

Paulo Cardoso Project partner · the build

Water and treatment followed the land instead of habit.

Readings arrived on their own, and the owner could see the state of the land the same day.

Rain data could be crossed with soil and sector readings. A manager could see that one sector was holding more water than another and irrigate accordingly instead of watering everything the same way. The same logic guided insecticide application.

The platform ran across three farms, with ten towers per farm and five to seven sensors on each tower.

IoT Mesh network Node.js AWS Offline sync Timestamp deduplication
Role
Ingestion APIs & offline sync
Partner
Paulo Cardoso
Model
IoT platform build
Status
3 farms deployed
Project partner Paulo Cardoso
A ticketing operations professional holding a handheld scanner at an event venue entrance
Ticketing · Platform modernization →

Ingresse

A ticketing platform made observable before its riskiest payment and ticket paths moved out of legacy PHP.

Read the case →
A dance instructor holding a phone in a quiet studio with an orange towel
Education · Payments →

FitDance

A risk-ordered WordPress rebuild with automated split payments and Mexico expansion.

Read the case →
All case studies 38 builds

Building field software? See a fixed-price build, an embedded engineering team, or our work in field services and events.

Next / Your build

Taking on new builds

Have something in mind?

Tell us what you're making. We reply within one business day, and you get a fixed price and a date in writing before any code.

Start a conversation More case studies