System design for physical AI: one agriculture data dashboard the operator fully owns

This is the engineering summary of an open reference architecture. The paper, its object model as JSON and the model register are free to reuse under CC BY 4.0: https://codeatoms.ai/farm-data-dashboard-pakistan/. The operator is described by class, never by name.

An agriculture and earth observation operator in Pakistan runs a federal productivity pilot. Its farm sensor feeds, its historical datasets and a big data and analytics repository sit in separate places, so nobody can drill from a season down to one field and one reading without a manual export.

Here is how that becomes one dashboard the operator owns outright.

1. Join three sources through two adapter families

A telemetry adapter reads the site sensor feeds. An integration adapter reads the analytics repository and the historical datasets. Both run as scheduled batch pulls. A streaming backbone was set aside, because the sensor cadences and volumes are not yet known and the prototype has to settle them first.

2. Twelve objects, one model

Site, field, crop cycle, sensor device, sensor reading, historical dataset, publication, dashboard user, access role, alert, report and data source connection. Which fields in this crop cycle raised alerts that nobody acknowledged is one traversal, not three exports.

3. Open source on the operator's own servers

The model is projected as a versioned schema in PostgreSQL with PostGIS. Nothing runs in a cloud the operator does not control. The source code and full intellectual property are handed over at the end.

4. No model is the right model here

The analytics are deterministic aggregations, drill down and reporting, reproducible from the operator's own data. The model register stays empty and no hardware is bought, because nothing in the requirement needs inference.

5. Prove the slope on one farm first

The rollout starts with a one farm integration prototype. It measures what each additional farm costs to integrate before the fixed budget is spent. Every alert is acknowledged by a named user under a role the operator assigns.

Full design, figures and the object model: the paper. DOI: https://doi.org/10.5281/zenodo.23186671

Designed on CodeNinja Praxis, the CodeNinja Atoms platform for designing physical AI systems. The object model imports into Hyper Ontology, which turns it into a living system. Load it yourself with the open hyper-ontology loader.