Review anonymised chest X-rays on your own schedule, from anywhere. Your verified readings train GeoRadX for the clinics that have no radiologist at all — and you're paid for every one.
It is the right question. Doctors have watched other professions be asked to train the systems that displaced them. You are owed a real answer, not a reassurance.
You are training your clinical assistant — not your replacement. And that is not a promise. It is a constraint built into what GeoRadX legally and structurally is.
GeoRadX is being registered as a diagnostic-support device. The terms of that approval forbid its use as a sole diagnostic instrument. A qualified clinician must read every case. That is not a policy we could quietly drop — it is the condition of the product's existence.
It flags regions that deviate from healthy anatomy and shows you why it flagged them. It does not name a disease, choose a treatment, or sign a report. The interpretation was never its to make. It hands a clinician a shorter list to look at — nothing more.
GeoRadX is built for clinics that have no specialist at all — where a film is read by a general clinician, or not read at all. The patient it helps was never going to receive a radiologist's read. There is no work there to take.
Every gain the model makes comes from a doctor's verified label. It has no other source of truth. It does not learn from the internet; it learns from licensed clinicians. It is permanently downstream of clinical judgment — a system that, by construction, needs doctors forever.
We will not tell you that AI never displaces anyone. That claim is not ours to make, and you would be right not to believe it. What we can tell you is what this particular system is — a tool that is forbidden from working alone, incapable of diagnosing, deployed where no radiologist exists, and dependent on doctors for every improvement it will ever make. You are not training the thing that takes your work. You are training the thing that extends your reach.
Asynchronous, remote, on your own schedule. No patient is waiting on you.
An anonymised chest X-ray, with a grid laid over it. Zoom, invert, adjust — the tools you'd expect.
Mark any finding, rank your confidence, and tick the exact cells where it falls. That spatial detail is what teaches the model.
Your read is recorded independently — you never see the engine's guess, and it never sees yours first.
The bar is deliberately high. The model is only as good as the doctors teaching it.
Applications take a few minutes. Verification takes up to three business days.
Reviewing is remote and asynchronous. It is training work — you are never responsible for a live patient's care through this network.
GeoRadX is early, and it is small. If you want a role where the thing you build actually reaches someone, that is the trade on offer.
Work directly on the DRF-MWU engine — geometric reasoning, edge inference, model compression. You'd need to be comfortable reading an unfamiliar architecture and extending it without breaking it.
Own the backbone: offline-first sync, encrypted patient records, the credit ledger, the reviewer pipeline. Hard problems with real constraints — devices that are offline for days, records we cannot decrypt.
A radiologist to own clinical validation, define the gold-standard case sets, and hold the line on what GeoRadX may and may not claim. This role has the power to say no.
Take NAFDAC approval through to clearance, and build the quality system that lets GeoRadX enter further markets without cutting corners.
Get GeoRadX into the clinics that need it. This is not enterprise sales — it is sitting with clinicians who have never used medical AI and earning their trust honestly.
If you can do something this project needs and it isn't listed, write anyway. Tell us what you'd build and why it matters. That letter is the application.
GeoRadX has one person building it. There is no team yet, no revenue, and no regulatory approval. The product has never been used on a real patient. Early roles are likely to involve equity rather than a market salary, and you would be taking a real risk on something unproven.
What is on the other side of that: a novel architecture that is genuinely yours to shape, a product with a clear reason to exist, and work that — if it succeeds — puts a diagnostic tool in the hands of clinics that have never had one.
If that trade is wrong for you right now, it is wrong, and there is no hard feeling in it. If it is right, we should talk.