Turning health documents into insights
Turning health documents into insights
Diagnostics / Consume app
Diagnostics / Consume app
A unified upload experience for lab reports, prescriptions, and scans
A unified upload experience for lab reports, prescriptions, and scans
00
00
Context
Context
00
Users carry their health history across a stack of disconnected documents — lab reports, prescriptions, and radiology scans, each in a different format and a different app or folder. Our product lets people upload any of these and get back plain-language, AI-generated insights: what the results mean, what to track, and what to do next.
I designed the end-to-end upload experience across all three document types, as a single coherent system rather than three separate features.
Users carry their health history across a stack of disconnected documents — lab reports, prescriptions, and radiology scans, each in a different format and a different app or folder. Our product lets people upload any of these and get back plain-language, AI-generated insights: what the results mean, what to track, and what to do next.
I designed the end-to-end upload experience across all three document types, as a single coherent system rather than three separate features.
01
01
The Problem
The Problem
01
A health record is only useful if it's understood. Most people receive a PDF or a photo of a report and have no way to interpret it. The core challenge wasn't the upload itself — it was designing a flow that could:
A health record is only useful if it's understood. Most people receive a PDF or a photo of a report and have no way to interpret it. The core challenge wasn't the upload itself — it was designing a flow that could:
handle three distinct document types (Rx, lab report, scan) that have different rules and outputs,
work for a family, not just one person. Users upload on behalf of parents, partners, and children,
and stay trustworthy while an AI processes sensitive medical data behind the scenes.
handle three distinct document types (Rx, lab report, scan) that have different rules and outputs,
work for a family, not just one person. Users upload on behalf of parents, partners, and children,
and stay trustworthy while an AI processes sensitive medical data behind the scenes.
02
02
Approach
Approach
02
Rather than building a bespoke flow per document type, I designed one shared journey and varied only what each document required:
Rather than building a bespoke flow per document type, I designed one shared journey and varied only what each document required:
Select member → Choose/confirm type → Upload from device → AI processing → Insight
Select member → Choose/confirm type → Upload from device → AI processing → Insight
This kept the mental model identical no matter what someone uploaded, while still respecting the differences — for example, lab reports capture extra metadata (report name, date, lab), prescriptions distinguish typed vs. handwritten, and scans surface organ-level inferences.
This kept the mental model identical no matter what someone uploaded, while still respecting the differences — for example, lab reports capture extra metadata (report name, date, lab), prescriptions distinguish typed vs. handwritten, and scans surface organ-level inferences.
03
03
Key design decisions
Key design decisions
03
A household-first model. Every upload begins by choosing who the document is for, with a first-class "Add a new member" path and a guided empty state for users who haven't added anyone yet ("You need to add a member in order to upload a report for them"). This made the product feel like a family health vault rather than a single-user utility
A household-first model. Every upload begins by choosing who the document is for, with a first-class "Add a new member" path and a guided empty state for users who haven't added anyone yet ("You need to add a member in order to upload a report for them"). This made the product feel like a family health vault rather than a single-user utility
Designing the wait, not hiding it. AI digitization takes time, so the processing screen reassures rather than blocks: "Feel free to close this screen — we'll handle everything in the background and notify you once done." Progress is shown both full-screen and as a compact mini-bar, so users can leave and the work continues.
Designing the wait, not hiding it. AI digitization takes time, so the processing screen reassures rather than blocks: "Feel free to close this screen — we'll handle everything in the background and notify you once done." Progress is shown both full-screen and as a compact mini-bar, so users can leave and the work continues.
04
04
The three flows
The three flows
04
All three share the same spine — select member → upload → process → insight — and diverge only where the document demands it.
All three share the same spine — select member → upload → process → insight — and diverge only where the document demands it.
Lab report
Lab report
Prescription
Prescription
Radiology scan
Radiology scan
05
05
Outcome
Outcome
05
The result is a consistent, reusable upload system spanning three document types and a full set of success, in-progress, and failure states — designed to scale as new record types are added.
The result is a consistent, reusable upload system spanning three document types and a full set of success, in-progress, and failure states — designed to scale as new record types are added.
4.7K
4.7K
Monthly avg uploads
Monthly avg uploads
06
06
Closing the loop (follow-on project)
Closing the loop (follow-on project)
06
Uploading and understanding a document is only half the journey. Once a user brings in their latest lab report or prescription, we leverages that data to power the next step — turning a passive record into personalised, actionable care:
Uploading and understanding a document is only half the journey. Once a user brings in their latest lab report or prescription, we leverages that data to power the next step — turning a passive record into personalised, actionable care:
Personalised lab tests — recommending the right follow-up tests based on the user's most recent results.
Relevant medicines — surfacing medications mapped to their current prescriptions and conditions.
Medicine & parameter interactions — flagging how prescribed medicines interact with each other and with the user's tracked health parameters.
Personalised lab tests — recommending the right follow-up tests based on the user's most recent results.
Relevant medicines — surfacing medications mapped to their current prescriptions and conditions.
Medicine & parameter interactions — flagging how prescribed medicines interact with each other and with the user's tracked health parameters.
Explored in depth as a separate project
Explored in depth as a separate project
Coming soon
Coming soon