teewai

Inspiration

I saw this guy on YouTube who had acne and made an app for himself and other people who had acne. I had a similar story with my teeth so I created teewai.

What it does

With the help of AI coupled with my researched system teewai analyzes your teeth and scores you based on several criterias derived from my studies and recommends measures to whiten your teeth.

How we built it

React , AI Vision, insforge, Three.js, React-Leaflet

Challenges we ran into

Honest shade analysis from a phone photo. The easy path is a scanner that always says "your teeth are getting whiter" — validating, but dangerously misleading. We rejected several shortcuts that produced inconsistent or inflated results before landing on an approach that reads a real shade from the photo and only accepts a full scan when every required angle is genuinely present.

Family-plan scan accounting. When a family member is granted scans, those credits live in the plan owner's pool — not the member's own account. Early grant logic silently failed because members couldn't see their scans until they left the plan. We had to make grants resolve through the same billing context the scan engine uses, so a gifted scan shows up immediately and correctly.

Privacy vs. convenience in photo storage. Fast image access and patient privacy pulled in opposite directions. We chose privacy: every photo is stored privately and only ever served through an ownership check, which meant rebuilding the entire photo pipeline around that constraint.

A long tail of edge-case bugs. Camera gestures, payment-method defaults, coupon rounding, and silent field-update failures each broke a real user flow. Each one taught us to trace the reported action end-to-end through the code rather than trusting the happy path.

Accomplishments that we're proud of

A dental engine, not a black box. teewai doesn't just ask an AI for an opinion — it runs a calibrated, clinically grounded layer that turns a photo into a real shade, a health grade, and a per-tooth breakdown, grounded in published dental indices. The same logic powers instant estimates and the authoritative result, so what you see is what's stored.

Scans that survive a dropped connection. A long AI analysis over flaky mobile data will disconnect. We made the scan persist immediately and finish server-side, so no one ever loses a scan — or the reward that comes with it — to a dead connection.

Privacy as architecture, not a setting. Private photo storage with ownership-checked access, no personal data processed by the scoring engine, and a verified-medical-professional badge with a medical-advice flagging and dispute workflow for the community.

A family plan that actually works. Scans are shared and transferable across a household, consumed in the right order, and visible to every member the moment they're granted — the billing matches how families actually share dental care.

Gamification that reinforces the habit. A ranked Smile ladder, unlockable secret smile traits, and milestone rewards — all delivered server-side, so the moment that keeps someone scanning through the weeks before visible change appears is never lost to a bad connection.