VitalLens

Inspiration

VitalLens was built around a simple idea: a wellness check-in should feel guided, accessible, and easy to understand without requiring extra hardware.

Most people already carry a phone with a camera, flash, motion sensors, and audio support, so I wanted to explore how those built-in sensors could create a lightweight wellness-only experience.

The goal was not to build a medical device, but to create a calm guided check-in that helps users capture a quick wellness snapshot using technology they already have.

What it does

VitalLens is a mobile-first guided wellness check-in app with two phone-based checks.

1. Finger-camera pulse estimate

The user covers the rear camera and flash with their finger, and the app estimates pulse from camera signal changes.

2. Breath motion check

The user places the phone on their chest or upper abdomen, and the app uses motion data to detect breathing-related movement.

After both checks, VitalLens creates a report with:

pulse estimate
breath motion result
signal quality
sample duration
AI wellness summary

The report uses IBM watsonx to interpret structured pulse and breath data and generate a short wellness-focused summary centered on:

session quality
signal reliability
breath consistency
practical next steps

VitalLens is not a medical device. It does not diagnose, treat, or make medical decisions.

How I built it

I built VitalLens with Next.js, React, TypeScript, and deployed it on Vercel.

For the pulse check, I used:

rear camera stream access
torch / flash support
finger-camera signal analysis
local signal quality tracking

The app also measures sample duration and signal stability so the result is not based on a random instant reading.

For the breath check, I used the DeviceMotion API to capture small chest-related movements during a guided 30-second session.

Because users may not be looking at the screen during this process, I added voice guidance to improve usability.

The Report screen combines local Pulse and Breath results, prepares structured telemetry, and sends it to a secure server-side API route connected to IBM watsonx.

IBM watsonx integration

VitalLens uses IBM watsonx through a server-side API route.

The app sends structured wellness-only data including:

pulse and breath results
signal traces
sample duration
basic signal statistics
breath motion consistency

Instead of simply repeating visible numbers, the watsonx summary interprets the quality of the session and generates a short understandable explanation of the results.

The AI summary is intentionally restricted to safe wellness language and avoids:

diagnosis
treatment recommendations
clinical claims
“normal” or “abnormal” medical wording

Challenges I ran into

The hardest challenge was making the experience reliable on mobile Safari.

Camera permissions, torch behavior, motion permissions, live video streams, and audio guidance all behave differently across mobile browsers and devices.

Another challenge was making the experience feel trustworthy. I wanted the breath waveform to reflect real motion data instead of decorative fake animation.

I also had to carefully handle IBM watsonx integration issues including:

environment variables
model availability
malformed JSON responses
fallback behavior when AI services fail

Accomplishments that I'm proud of

I’m proud that VitalLens combines multiple built-in phone capabilities into one guided experience:

camera
flash
motion sensors
voice guidance
local signal processing
IBM watsonx summaries

I also designed the app so local wellness results still work even if the AI summary becomes unavailable, making the experience more reliable and resilient.