AI Vision Capture for Telehealth Mobile App

Adoption of the telehealth app's vitals logging was weak, and the obvious read was that manual entry was too tedious. It wasn't — the real constraint was device coverage. Each Bluetooth device brand required its own native SDK integration, so patients whose devices weren't supported had no path in at all, and manual entry was producing transcription errors that left care coordinators with unreliable data. I proposed photo capture as the way around both: patients photograph the device display, readings auto-fill, manual entry stays as a fallback.

What I did

• Owned the feature end to end — designed the screens, reworked the existing entry flows around the new capture method, led the development team, and ran acceptance testing
• Specified the Azure AI Vision integration, including partial-read handling and a patient verification step before any reading is saved
• Designed server-driven brand-to-connection-method mapping so adding or changing a device brand doesn't require an app release
• Removed the need to integrate an SDK for each new device brand — an estimated $30–40K saved and a shorter path to supporting new hardware
• Authored release documentation distributed across Product, Operations, Clinical, Support, and Leadership