GeoRadX is offline chest X-ray diagnostic support built on the DRF-MWU engine. It runs on the hardware a clinic already owns, calibrates to the population it serves, and keeps every patient's data on the device that read it.
GeoRadX reads chest X-rays for signs of disease in seconds. It flags findings to the exact region of the chest they sit in, shows the measured reasoning behind each one, and leaves the final call with the clinician. The read itself runs entirely on the device — no internet, no cloud, no wait.
Around that engine sits a full clinical layer: patient records, follow-up tracking that compares a patient's films over time, and a copilot that explains a read in plain language. These connected features reach our cloud, while the reading never does. A clinic manages its own accounts and access, and no patient image ever leaves the device — only the engine's numeric measurements are ever synced, and an X-ray cannot be rebuilt from them.
An X-ray is often taken far from the nearest radiologist — in a rural clinic, a screening camp, a mobile unit — where a read can take days to come back, if it comes at all. And the diagnostic AI meant to help is trained overwhelmingly on North American and European populations, then shipped everywhere unchanged, carrying a bias into exactly the settings that can least afford it.
GeoRadX was built for the other setting: low power, low bandwidth, and the population actually in front of the machine.
Diagnostic intelligence and patient-record management that run entirely on existing, low-spec hardware — without internet, and without exposing anything to cloud risk.
Runs offlineCell-by-cell comparison of a patient's films over weeks or months, giving clinicians objective evidence of whether disease is clearing, holding, or spreading.
Internet requiredAn engine that calibrates to the demographic and anatomical baseline of the region it serves, so diagnostic accuracy holds for the people actually being screened.
Periodic connectionAn assistant that translates a read into plain language and lets clinicians interrogate a case — questioning a region, comparing against the wider population, exploring the engine's findings.
Internet requiredGeoRadX starts where the need is sharpest and the infrastructure thinnest — bringing a read to the patient rather than waiting for the patient to reach a read.
NGOs and public-health programmes running screening in the field — multiple teams under one organisation, moving between locations, with the engine re-calibrating to each as they go. This is where GeoRadX begins.
Rural clinics running a single device, larger facilities augmenting imaging they already own, and hardware makers integrating GeoRadX into what they build. The same engine, a different footprint.
Everything GeoRadX does rests on a proprietary architecture called DRF-MWU — a spatial reasoning engine that is coherent by construction, empty of parameters at baseline, and light enough to run at the edge. It is held as a trade secret, and the working product is how we let it speak.
Thoracic pathology on frontal chest films — the first domain the engine is deployed into.
Development of a mammography domain begins in January 2027, extending the same architecture to breast imaging.
Because the engine takes on a domain from a few hundred examples, each new area is a calibration rather than a rebuild.
A medical-AI company building offline, population-adaptive diagnostic imaging for the settings that conventional tools overlook.