Case study
Offline-first Android app shipped for a clinical health platform. Logs accurate even underground.
The client applies machine learning to epilepsy treatment. The patient app is the clinical data layer. It had to log seizures and medication events in real time, in any location, with or without signal.
Healthcare / Digital Health · Early-stage, venture-backed · North America
The challenge
A seizure doesn't wait for a signal
The client's platform applies machine learning to epilepsy treatment. The model needs accurate patient data to work. Patients log seizures, side effects, and missed doses. Doctors use those logs to adjust medication regimens. Gaps or delays in the data produce worse treatment outcomes.
Patients don't have seizures only in places with strong signal. A log entered hours after the fact is less accurate than one entered immediately. The app had to accept event logs with no connectivity and sync them automatically when signal returned. No manual sync. No data lost. No instruction to the patient.
Medication reminders were equally critical. Epilepsy medication is time-sensitive. A reminder that fires late or not at all is a clinical risk. Standard Android notification behavior is not reliable enough when the OS is in battery-save mode or when the app is in the background. The team needed a custom notification system that guaranteed delivery regardless of device state.
Doctors needed to see the full picture. That meant graphical summaries of seizure frequency, side effects, and missed doses over 7, 30, and 180-day windows. Patients needed a calendar showing what happened on any given day and the ability to edit historical logs when the doctor asked for corrections.
Articles and educational content also needed to be available without connectivity. The app had to surface relevant articles in multiple places based on what the patient was doing.
“The data is only useful if it's captured at the moment it happens. Anything that gets in the way of that has clinical consequences.”
The approach
Custom notification engine built offline-first from the ground up.
Wednesday assigned a team of Android engineers to review the clinical requirements and build an implementation plan before work began.
Medication reminders. Standard Android notifications don't guarantee delivery. The team built a custom notification engine using Android's AlarmManager for precise timing, foreground services to schedule the next reminder automatically after each fires, and WorkManager to satisfy Android's background processing and battery optimization requirements. Reminders fire on schedule regardless of whether the device is in battery-save mode or the app is not in the foreground.
Offline-first logging. Realm was used as the local database. When a patient logs a seizure or side effect, the record is written to the device immediately. Sync happens in the background when signal is available, without any action from the patient. Changes made while offline are automatically reconciled once connectivity returns.
Health history views. The team built a calendar showing all log events for a selected day, with the ability to edit past records. A graphical summary at 7, 30, and 180-day ranges gives patients and doctors a view of seizure frequency, side effect patterns, and missed dose rates at a glance.
Offline articles. Contentful was used as the content management platform. The team built a conversion layer that stores article content locally in a format the app can read without connectivity. Articles are surfaced as contextual suggestions throughout the app based on patient activity.
Release automation. The team created separate app configurations for each environment, so development, staging, and live builds can all be installed on the same device simultaneously. GitHub Actions and AppCenter handle the build and release process automatically for both platforms.
“They really cared and felt like an extension of our team. The quality of the work was top notch, and they were receptive to shifting priorities.”
The results
Logs are accurate. Reminders fire on time. No data lost.
The app logs seizures and side effects from anywhere, without signal. The data syncs automatically and accurately. Doctors get a complete picture. Patients get 180-day views of their health history, graphically, with day-level detail available by tapping any date on the calendar.
Medication reminders fire on the scheduled time regardless of device state. Battery-save mode does not suppress them. The clinical integrity of the reminder system holds.
Articles are available offline throughout the app and surface as contextual suggestions based on patient activity. Contentful manages the content. The app handles the offline delivery without any patient-facing complexity.
Both platforms build and release automatically. The clinical team can update the app on their schedule without a manual release process.
“They delivered within a short period of time and met all our expectations. The process was smooth from start to finish.”
ROI
An unreliable patient app is a liability. Gaps in the clinical data produce inaccurate treatment recommendations. The cost of a wrong recommendation is not measured in software rework.
Run the numbers
See what these results would look like for your team size and budget.
“They really cared and felt like an extension of our team. The quality of the work was top notch, and they were receptive to shifting priorities from our end.”
Founder — Digital health platform
Next step
Facing a similar offline-first mobile build?
30 minutes with an engineer. Bring your current setup and your deadline. You leave with a squad shape and a written burn estimate.