Skip to content

SORIdent

An app that replaced a dental clinic's paper records, reading years of handwritten sheets with AI and running the calendar, treatment plans and sterilization tracing.

Industry
Dental clinic
Scope
Design and build of the clinic's internal app
Platform
Next.js on Cloudflare Workers
Duration
2 weeks
Location
Bucharest, Romania
Design
PixelMakers
Build
PixelMakers
A handwritten SORIdent consultation sheet next to the app's review screen, with the patient's name, birth date, phone and address read from it

SORIdent is Dr. Sorin Dragnea's dental clinic in Bucharest. Its records lived on paper: a consultation sheet for every patient, a handwritten treatment log, and a sterilization annex listing every steam pouch used. We built the clinic its own app to replace them, used all day on the desktops and tablets in the treatment rooms.

Paper in, records out

Years of patient folders could not be retyped. The team photographs a folder on a tablet, every page in any order, and the app takes it from there. Each page is read by AI, sorted by type and page number, and turned into the fields of a record: identity, health questionnaire, allergies, tooth shade, and every line of the treatment log with its date and tooth.

A paper consultation sheet and a handwritten treatment log
Where it started: the consultation sheet and the treatment log. Shown with an invented patient.

Nothing reaches a record on the AI's word. The review screen puts the photo of the page next to the fields read from it, tints anything hard to read, and suggests which patient the folder belongs to, calling a match sure only when the name and a phone number or birth date agree. A doctor approves every page before it is saved.

A handwritten treatment log next to the rows the AI read from it
A treatment log read line by line, with the tooth written as 36 stored as 3.6.

The clinic's day

The calendar shows the week for every doctor and treatment room, synced with the clinic's Google Calendar every five minutes. A booking opens in a side panel with the patient's phone, its status and a link to the record. A double booking can't happen: the check and the write are one database statement. Patients with an email address get a reminder the day before.

Today and Tomorrow list every appointment with the patient's medical alerts, colored by condition, so an allergy or an anticoagulant is in view before the patient sits down.

A booking opened from the calendar, with its status, doctor, room and reason
Every booking opens from the calendar. Shown with invented patients.

One record per patient

  • A tooth chart in FDI notation: tap a tooth for its history or to log a treatment.
  • Treatment plans that are also the quote, priced from the clinic's own list and ticked off line by line as the work is done.
  • Agreements and plans signed on the tablet, sealed with a SHA-256 fingerprint and kept as locked PDFs.
  • X-rays and CT reports in a full-screen viewer, beside the notes and the treatment log.
  • Sterilization tracing: each pouch label links a visit to its autoclave cycle, and each cycle to every patient it was used on.
A patient's tooth chart with planned and treated teeth marked
The tooth chart, with what's done and what's planned.

Built for medical data

Patient data stays in Cloudflare's EU jurisdiction and is never cached. Every read and every change to a record writes an audit entry. Staff accounts are invite-only, with two-factor sign-in or passkeys, and a shared tablet locks itself after 30 minutes idle. AI readings run with logging and storage turned off.

The challenge

A clinic that runs on paper can't stop seeing patients to go digital. The app had to fit the way the team already works, read their handwriting, and take over the old folders without anyone retyping them, while handling health data with the care it needs.

Our approach

We started from the paper. Every field on the sheets became a field in the app, in the same order, so the screen reads like the sheet the team knows. We agreed the plan with the clinic first, built it in phases, and shaped it with the team's feedback as they used it.

Results

Two weeks after the plan was approved, the clinic had its full first version: patients and records, the calendar, treatment plans and quotes, signing on the tablet and sterilization tracing, with the paper archive coming in through the AI reader at about two cents a page.

14 days
From the approved plan to the full first version
$0.02
AI cost to read one paper page
100%
Pages approved by a doctor before they reach a record

Stack

App
Next.js 16 on Cloudflare Workers
Data
Hosted by Cloudflare in the EU
AI reading
AI document reading through Cloudflare AI Gateway
Integrations
Google Calendar sync, email reminders
Interface
Tailwind CSS and shadcn/ui