Hackathon
MemoryCare
Next.js Tailwind CSS Supabase FaceAPI Gemini Live
Overview
Built at TerraHacks (24-hour hackathon). Prosopagnosia is a condition where people struggle to recognize faces, even people they know well. Combined with memory loss, this can be really isolating. MemoryCare lets users add photos of loved ones with notes about them, then uses real-time face detection to identify who's in front of the camera and starts a personalized Gemini Live conversation to help recall memories about that person.
What I built
- Real-time face detection and recognition via FaceAPI
- Gemini Live for natural, conversational memory recall
- Privacy-first: face data is processed client-side, not sent to servers
- Personalized memory profiles for each recognized person
- Built and shipped in under 24 hours
- Supabase for minimal, secure user data storage
What I learned
- Integrating multiple AI services (FaceAPI + Gemini Live) into one coherent UX
- Real-time video processing in the browser without performance issues
- Designing with user privacy as the primary constraint, not an afterthought
- How to scope and ship something meaningful under extreme time pressure