Inspiration

Skincare is sold on marketing claims, and objective skin imaging costs thousands of dollars in a clinic. A phone camera is good enough when the lighting is controlled. So we built the loop: measure, change one thing, measure again.

What it does

Scans your face and scores 6 skin metrics (0–100)
An ML detector maps and counts every acne lesion
An AI agent combines the scan with your sleep, diet and UV index and writes today's plan, with the reason for each item
Tells you whether to keep, stop or swap each product you use
Tracks your skin trend over time

How we built it

SwiftUI iOS app + FastAPI backend with an OpenCV scoring pipeline. Acne detection: YOLO + MediaPipe. Agent: gpt-5-mini with structured outputs, falling back to a rules engine if it fails.

Challenges

We only had white light, so metrics we can't measure yet are labeled "estimated". Acne detection on a full photo missed almost everything, so we crop to the face first.