ScrybeDoctors at Kasr Al Ainy copy handwritten admission sheets into MRM, the records system, by hand. Scrybe, an AI app I programmed for the ward PC, does the copying and stops short of saving.
01 / 05A doctor photographs the sheet with a phone, and the photo comes over to the PC by cable.
02 / 05The patient's name is masked before any model sees the page. Scrybe never stores it and never sends it.
03 / 05Gemini and Groq each read the handwriting separately. If their readings differ, the field is flagged and both versions are shown side by side.
04 / 05The doctor goes through every field, fixes what is wrong and confirms the rest.
05 / 05Scrybe types the confirmed fields into MRM. You press Save, because the app never will.
RotaMateIn RotaMate, my AI-supported shift app, a Duty Assistant answers interns' questions from the published guides. The booking side runs on rules I wrote, and this is how they share out the cover.
01 / 03Interns pick the shifts they want. Left to first come, first served, a few keen people end up holding most of the fortnight.
02 / 03Then the rules count everyone's shifts and hand cover from the busiest to the least busy, never onto the day after a night. Each intern sees every change that touches them.
03 / 03The same fortnight, shared out evenly.
Internal Medicine recordsInternal Medicine records is the department's records app. The part I'm building now lets a doctor say the note instead of typing it.
01 / 04The doctor opens the patient's page and types the name and file number by hand. The AI never hears either of them.
02 / 04Then they say the rest out loud, moving between Arabic and English as they like, and Gemini turns it into the department's template.
03 / 04Every field it fills stays marked. A blood pressure of 1500/90 can't be right, so it gets flagged and left empty.
04 / 04The doctor checks each mark and signs. Only then is it saved, logged as AI-assisted, and the recording deleted.
Alumni systemProf. Dr. Ehab El Refaee and I came up with the alumni system, and I programmed it with AI coding agents. It finds where the faculty's graduates went and what they've published.
01 / 03The agents worked through a plan one task at a time, and more than 500 tests check what they wrote.
02 / 03There's no AI inside the app itself. A fixed rule scores each match: half for the name, a quarter for a Kasr Al Ainy link, the rest for specialty and timing.
03 / 03A strong score only marks a match. It still waits until a person says it's the same doctor.