
Ucentrix: A SaaS Product Developed for a Medical Clinic
Track & Share Individual Medical Records
Overview
I was the only product designer on Ucentrix, a multi-role health records platform. I took it from first wireframes to a live product in 20 months. Clinics use it to connect with patients, patients keep their full medical record in it, and AI reads their lab reports so they don’t have to type every value
Multi-role health records SaaS Web Application for clinics, doctors and patients
20 months
Product Designer, 2 Frontend Developers, Backend Developer, Client

Responsive web and mobile, built in Flutter
14 workflows and 114+ screens across a patient portal and a doctor portal
10-section medical profile, 4+ data-heavy lab report types, AI data extraction, role-based access
The Problem
A wrong number in a health record is worse than a missing one
Health records on their own can be confusing. When all your records are dispersed, it’s difficult to keep track of your health.
The clinic needed one place where patients and doctors share a complete health record. That record is large, varied and mostly locked inside PDFs.
Three things made it hard:
Lots of data, in many shapes
The medical profile alone has 10 sections. Every lab report type (bloodwork, DEXA, VO2 max, imaging) has its own fields, units and layout.
Typing is slow, AI is fast, and AI can be wrong
Reading a report with AI saves time, but a misread value saved silently is the worst outcome.
Two roles, one record.
Patients own the data. Doctors need to scan it between appointments. Each needs a different view, and the patient decides who sees what.
The design question: how do we make a large, sensitive dataset fast to enter, easy to scan and safe to trust, for two very different users❓
Who it's for
Two groups use the same platform for different jobs, and both need to trust that the data is right and find it fast.
User
Individuals/ Patients
What they need
Want their medical history, medications, allergies and results in one place, and a simple way to share them.
What gets in the way
Today they get scattered PDFs and units they can’t read.
Healthcare
Professionals/ Doctors
Want to review a patient quickly, often between appointments.
Legacy systems give them cluttered screens and make them hunt for the one value that matters.
Because the two portals share one product, I set a rule early: the same components and patterns serve both, so moving between roles never feels like switching apps.

The system at a glance

My Process
I worked directly with the clinic and the Flutter developers, in this order:
Requirements Gathering
Interview the business owner for their requirements, competitor product analysis, end user requirements specification
Map Workflows
Split the product into 14 workflows (8 patient, 6 doctor).
Wireframe
Agree on structure in low fidelity before any visual design.
Build the system
One Figma library of buttons, fields and cards for both portals.
Design
114+ final screens for web and mobile.
Validate
Present to the client, receive feedback.
Hand Off
Development-ready designs, plus advice on what screens can’t show, like screen reader labels.

Key Challenges
Challenge 1:
Making AI-extracted data trustworthy
I designed a human-in-the-loop review:
AI reads the uploaded report and suggests values, and a person approves every one before anything is saved. Five decisions shaped it:

The same pattern covers bloodwork, DEXA and VO2 max, so users learn it once. Uploaded and manually entered reports end on the same review screen, which means one pattern to learn and one place where data is approved.
Challenge 2:
Making dense medical data scannable
Health records are dense by nature, so I made the layout do the sorting. Six patterns carry the whole product:
Challenge 3:
Two roles, one record, patient-controlled access
All 14 workflows read and write the same record, and both portals use the same components, so a doctor reviewing a report sees the layout the patient used to add it. Sharing raises one question on every screen: who can see this? Patients own the record, so they control access:
On the doctor’s side the same rules decide which patients and reports appear, so a doctor only ever sees what a patient has shared.
Built to Scale
A component system for one designer and 114+ screens

With no second designer, consistency had to come from the system, not from memory.
Design once, reuse everywhere: Buttons, fields and cards are shared by every workflow.
2. Repeatable page patterns: Every report follows the same structure (source, grouped fields, confirm bar), so a new report type reuses a pattern instead of starting from scratch.
Clear signals: Orange is reserved for primary actions and things that need attention.
Faster handoff: Developers build each component once and reuse it.
Accessibility
I designed to WCAG 2.1 AA, because a health record has to work for everyone. These are the decisions that carry it, and where to see them above.
Colour is never the only signal (1.4.1): Allergy severity, condition status, connection and access level all pair colour with a text label: “Severe”, “Connected”, “Full Access”. See Challenge 2, screen 6, and Challenge 3, screen 3.
Nothing is saved until a person confirms it (3.3.4): AI suggestions are accepted one value at a time. Confirm and Save is the only way data enters the record, and Discard is always one click away. Doctor access is reviewed before an invite is sent and can be changed at any time. See Challenge 1.
Structure a screen reader can follow (1.3.1, 2.4.6): Every value sits in a labelled column with its unit and reference range. Fields have visible labels, and sections have clear headings. Buttons say what they do: “Add Allergy”, not “Add”.
More than one way to find a record (2.4.5): Side navigation, breadcrumbs, and search, filter and sort on every report list.
Consistent everywhere (3.2.3, 3.2.4): Both portals share one sidebar and one component library. Every action menu lists View, Edit, Download and Delete in the same order.
Screen reader labels in the handoff (4.1.2): I specified what assistive tech should announce, such as “Change suggestion for Hemoglobin” instead of “button,” and “Current 12.8, suggested 13” instead of an arrow.
What I’d fix next: Measuring the final screens showed the orange primary buttons reach about 4.1:1, short of the 4.5:1 AA needs, and two small status labels fall short too. Darkening the orange is the first change. I’d also add visible focus states to the component library and a table view for the trend charts.













