Building a consumer AI app right now is wild. Building one that relies on biometric data? That adds a massive layer of complexity, especially when you want to ensure total user privacy.
Most "lookalike" apps out there rely on creepy web scrapers or basic reverse image searches. I wanted to build something entirely different: a 100% closed, opt-in biometric network where the user owns their face data.
So, I built DopplGrid. Here is a look at the stack and the architecture behind it.
The Stack & Architecture
I wanted to get the web application running flawlessly before wrapping it for the native iOS and Android releases. The frontend is built in React and heavily styled with Tailwind CSS to keep the UI clean, fast, and responsive across devices.
For the backend and secure data vault, I am relying on Firebase. The core of the app is a biometric matching engine. Instead of scanning colors or pixels, it maps 128 unique points of a user's facial geometry.
The Privacy Constraint (No Web Scraping)
The biggest architectural rule I set was zero open-web scraping. The app acts as a personal photo radar, but it only tracks photos uploaded within the DopplGrid network. It only starts working after a user explicitly chooses to securely enroll their face. This ensures users can safely track their digital footprint without their data being aggregated from external sources.
The Features (The Fun Part)
Once the secure vault was operational, the matching algorithm opened up some really fun use cases.
Global Matching: Users can scan the network to find your doppelganger globally based on exact geometric similarity.
Family & Friends: I added a similarity tool so users can compare their facial features directly with friends or relatives to see the exact percentage of how much they look alike.
Looking for Feedback
We just pushed the initial React web build live. If you are interested in AI tools, privacy-first architecture, or just want to see how accurate the matching engine is to find people who look like you, I would love for the dev community to test it out.
Check out the onboarding flow at DopplGrid and let me know what you think of the UI/UX!