Inspiration

Over 7.2 million Americans live with Alzheimer's disease, including my grandfather. Losing track of medication, keys, or familiar places can quickly become a safety risk. We wanted to build an AI companion that helps people remember and stay independent.

What it does

Findr is an AI memory assistant powered by Meta Ray-Ban smart glasses. It continuously builds a searchable memory of what the wearer sees, allowing users to ask questions like, "Where did I leave my keys?" During emergencies, Findr instantly alerts caregivers with recent context via SMS.

How we built it

Findr connects Meta Ray-Ban smart glasses to a FastAPI backend through an iPhone bridge app. Computer vision filters redundant frames, Claude Haiku converts meaningful moments into searchable memories, ChromaDB and SQLite power semantic search, Whisper handles speech recognition, and Twilio delivers caregiver alerts.

Challenges we ran into

The hardest challenge was a regression in Meta's Wearables SDK that prevented reliable camera streaming. We isolated the issue, deployed a stable workaround, implemented automatic session recovery, and improved our AI prompts to generate accurate, searchable memories while avoiding hallucinations.

Accomplishments that we're proud of

End-to-end AI memory system running on real Meta Ray-Ban smart glasses
Searchable visual memory with grounded AI responses
Real-time caregiver alerts with contextual memory summaries
Reliable pipeline optimized for accuracy, speed, and cost

What we learned

We learned that building reliable systems on real hardware is often harder than building the AI itself. We also learned that careful prompt engineering is essential for creating trustworthy, searchable memories.

What's next for Findr

Next, we plan to improve object and text recognition, personalize the AI for each user, support additional wearable devices, and expand caregiver features. Our vision is to make AI a trusted companion for people living with memory impairment.