Patient screening kiosk

Inspiration

I work in healthcare and see daily how automating certain tasks could reduce workload and increase efficiency. Also we are currently working on a pharmacy healthcare app and that got us thinking of this

What it does

Patients check in via ID/phone lookup against a local SQLite DB, a single photo is captured on-device, an AI classifier judges it under a hard timeout, identity is confirmed field-by-field, if kiosk shows fail or flagged, this triggers a see the front desk never a silent pass.

How we built it

C++17 on QNX SDP 8.0, running on a Raspberry Pi 5. Direct camera_api.h calls for one-shot camera capture (no OpenCV), SQLite for patient records and logging, and a detached-thread fail-safe wrapper that hard-times-out the AI classifier so it can never hang the app.

Challenges we ran into

Accomplishments that we're proud of

A fully working, hardware-verified pipeline in one night — real camera capture, real fail-safe behavior (timeout and low-confidence both correctly route to FLAGGED), and zero cloud calls anywhere in the screening path.

What we learned

The fail-safe logic matters more than the model behind it — designing a swappable AI interface up front meant a incomplete PyTorch integration didn't block a working, demo-ready system.