UNIVERSITY OF BRASÍLIAMapa
Safra
Draw your production units and see, month by month, which plots were productive.
A University of Brasília project. An artificial intelligence model reads public satellite imagery and shows where production happened inside the perimeter you draw. Try it on your own property.
Looking at the field is expensive — and it costs again the next month
No plot outlines
You have the farm's registry number, not the outline of its plots.
Relies on declarations
What was planted in each area depends on what someone declared.
Late information
By the time the information arrives, the harvest is already over.
Expensive field visits
Going to the field is expensive, sample-based, and repeated with every new question.
Cost doesn't scale
The cost of the traditional method grows with hectares × number of inspections. The cost of satellite imagery grows with neither.
Plot, not perimeter
The productive plot, not the farm's entire perimeter
From perimeter to monthly answer
You draw the perimeter
The land is split into cells
The satellite follows the season
You choose the answer's resolution
A source that doesn't depend on anyone reporting
Sentinel-2 imagery is public, free and refreshed every five days. The image was never what was missing. It was turning it into an answer in a useful format.
Draw the perimeter
Draw the property's perimeter on the map.
Get the plots
Get the productive plots identified inside it.
See it month by month
See month by month which ones produced and what they produced.
Where the answer becomes a decision
The value isn't the map. It's what stops depending on someone driving out there to find out.
- 01
Visit triage
The satellite shows where it's worth sending someone. Inspection stops being a lottery and becomes a choice.
- 02
Declaration check
A second, independent source on what was declared — without having to confront anyone.
- 03
A view of the whole portfolio
Covering every area every month stops being an ambition. Sampling is expensive precisely because it requires choosing where to go.
Trained and evaluated on 1,754 real agricultural fields in Bahia, over a full crop year.
Most answer per harvest. We answer per month.
An answer that arrives at the end of the harvest is not information, it's a historical record. Credit looks at the cycle. Insurance claims look at the month of the event. Oversight looks at the year.
Start with one region
The model was validated in western Bahia over a full crop year. For other regions, we run a validation pilot: you bring areas you already know, we return the monthly reading and a measured-accuracy report. We prefer to state the limit rather than sell beyond it.