Sensor telemetry to prescription maps, one pipeline.
Tell us about your sensors and your fields. AI-native engineers build the edge firmware that reads them and the PostGIS/Mapbox pipeline that turns readings into prescription maps — one team, both skills, in your timezone.
Start with where precision ag actually breaks
It breaks at the seam between the device and the map.
Two vendors, two data models
Mismatched schemas, and telemetry that will not line up with your field geometry without someone reconciling it by hand.
It works in the demo, stalls in the season
A pipeline that handled a few test plots buckles the first time a full season of readings lands on it.
Firmware plus geospatial is two scarce hires
Two thin markets, and the platform is worth very little until you have both.
Every one of those lives at the same seam. Close the seam and none of them exists — which means one team has to own both sides of it.
So here is the pipeline, end to end
Field-to-Map
One pipeline from sensor to map, rather than two vendors bolted together at a schema neither of them controls.
The device reads and preprocesses at the edge
Readings are interpreted on the board — on-device ML wherever it earns its place — so what leaves the device is already worth sending.
Telemetry moves over MQTT, store-and-forward
Fields have dead spots. Nothing is dropped — the device holds the readings and catches up when the signal returns.
PostGIS joins it to zones and imagery
Readings land in PostGIS, get joined against zones and imagery, and come back as variable-rate prescription maps.
One pipeline, one team, one schema — the integration seam that usually eats a season is simply not there to break. And we build it AI-native throughout: the model runs on the board when the decision has to be instant, and in the pipeline when it has to see the whole field at once.
So here is what we have shipped
Hawk-Eye and NESS Fertigation, in production.
We built the Hawk-Eye crop-monitoring platform for Itri Corporation — continuous sensor streams, stored at scale, surfaced through dashboards — and the NESS Fertigation IoT portal with a custom ThingsBoard. AI-native engineers, fluent in on-device ML and agentic tooling: firmware and geospatial from one experienced nearshore team, one to four hours off US time.
“I could say what I needed, and ByteNana turned it into code.”
Tech stack
- ESP32 / FreeRTOS
- On-device ML
- MQTT
- PostGIS
- Mapbox GL
- QGIS
No integration seam between device and map.
Turn sensor telemetry from the field into prescription maps your users act on — built by one team that owns both the edge firmware and the geospatial pipeline.
So bring us the sensors and the fields.
What you measure, what the hardware is, and what the prescription map has to say — straight to the engineers who would build both halves. We'll map the shortest path from field sensor to a map your users act on.
We reply within one business day. We use your details to respond and schedule your call, as described in our Privacy Policy.