Diagnose diseases, smartly.
AMDR SYSTEM
A group of professors and doctors needed a platform to make use of academic disease-diagnosing knowledge alongside practising doctors' own expertise, without turning that judgment over to a black-box model.
Patients enter symptoms through a guided intake flow; a rules-based relevancy engine scores each candidate diagnosis against the symptom set and surfaces ranked matches with a visible percentage, not a single verdict. Every prediction routes into a doctor's review queue before it reaches the patient, with separate dashboards for patients, on-call doctors, and referred specialists.
Deductive, not black-box
We built a weighted symptom-matching model the academic team could inspect and tune directly, instead of a machine-learning model whose reasoning a doctor couldn't audit, critical for a system making medical suggestions.
Human-in-the-loop by default
The system never returns a diagnosis directly to a patient; every predicted match is queued for a doctor to confirm, adjust, or refer onward first.
Role-specific dashboards
Patients, doctors, and referred specialists each get a purpose-built view rather than one shared screen gated by permissions, so every role only sees what it needs to act on.
What shipped is a diagnostic aid doctors can actually trust: transparent enough to override, structured enough to speed up the parts of triage that don't need a specialist's full attention.
Have a project in mind? Let's talk.
START A CONVERSATIONEdit the code, watch it run.
Change the function on the left: your letter's background updates instantly. No reload, no waiting.