Agriculture. AI Systems, Mobile.
Farmers get a diagnosis from a phone photo, offline.
- Client
- AgriTech client, Pakistan
- Industry
- Agriculture
- What we built
- AI Systems, Mobile
- Timeline
- 6 weeks, 2024
Before
An image classification model running entirely on the device, returning a diagnosis across six-plus diseases without a network connection.
Farmers were waiting days for a veterinary diagnosis, in conditions where connectivity cannot be assumed. A cloud-based tool would have failed exactly where it was needed most.
What we built
01
The decision that shaped it
We took the network out of the loop entirely. Running inference on the device means the constraint that would have killed the product, patchy rural connectivity, stops being a constraint at all. That decision set the model size budget, and everything else was engineered to fit inside it.
02
What we engineered
An image classification model covering six-plus diseases, running entirely on the handset. Capture is a phone photograph; the diagnosis returns in seconds with no round trip.
03
Classification across six-plus diseases
On-device inference, no internet required. Phone-camera capture workflow.
04
What leaves the system
Diagnosis moved from a multi-day wait to the moment of inspection, in the field conditions where the decision actually gets made.
01
The decision that shaped it
We took the network out of the loop entirely. Running inference on the device means the constraint that would have killed the product, patchy rural connectivity, stops being a constraint at all. That decision set the model size budget, and everything else was engineered to fit inside it.
02
What we engineered
An image classification model covering six-plus diseases, running entirely on the handset. Capture is a phone photograph; the diagnosis returns in seconds with no round trip.
03
Classification across six-plus diseases
On-device inference, no internet required. Phone-camera capture workflow.
04
What leaves the system
Diagnosis moved from a multi-day wait to the moment of inspection, in the field conditions where the decision actually gets made.
- Classification across six-plus diseases
- On-device inference, no internet required
- Phone-camera capture workflow
What changed
Farmers get a diagnosis from a phone photo in seconds, offline.
Diagnosis moved from a multi-day wait to the moment of inspection, in the field conditions where the decision actually gets made.
Have a process shaped like this one?
Describe it in plain language. We will respond with a proposed architecture, a first milestone and the key risks.
We respond within one business day. If we are not the right fit, we will tell you.