Memerai

Data collection demo case
Data collection device
Simple version overall workflow
Comprehensive version overall workflow
Data collection workflow
Backend workflow
Memory gallery
Image memory training page
Home page
Launching page

Inspiration

Millions of families watch loved ones lose memories to Alzheimer’s. We wanted to create a warm, AI-powered companion that not only helps patients recall moments but also strengthens their connection with family. Our inspiration came from the idea that technology should feel human — comforting, familiar, and empathetic.

What it does

Memerai is an AI memory companion that helps early-stage Alzheimer’s patients remember faces, events, and emotions.

Uses face recognition to identify familiar people in photos.
Converts daily conversations and images into structured “memories.”
Provides personalized recall training through friendly voice chats and reminders.
Allows caregivers and family members to view progress through a connected dashboard.

How we built it

Hardware: Used a Raspberry Pi with a camera and microphone module to capture daily moments automatically.
Designed and 3D printed a lightweight CAD-modeled case that can be comfortably worn around the neck for continuous capture.
Frontend: Built with SwiftUI for iPadOS, featuring a clean and accessible UI.
Backend & AI agents: Multi-agent system using LangChain + GPT-4o for recall, summarization, and emotion analysis.
Face recognition: Gemini Vision + OpenAI Embeddings for photo tagging.
Voice interaction: OpenAI TTS for natural, warm speech.
Database: Pinecone for memory embeddings and timeline retrieval.
Notifications: iPadOS local reminders at 10 AM to encourage daily engagement.

Challenges we ran into

Hardware: Dealing with broken camera modules and GPS. Software: Configuring Swift privacy permissions on Xcode and secure API connections. Designing a user-friendly UI that feels intuitive for elderly users. Improving face recognition precision across lighting and pose variations. Managing data privacy and runtime stability across multiple APIs.

Accomplishments that we're proud of

Built a working prototype with voice conversation, face recognition, and daily memory training. Designed a visually calm interface for elderly users. Integrated OpenAI TTS for expressive, natural-sounding speech. Created an early version of a personalized memory-training curve based on user responses.

What we learned

How to coordinate multiple AI agents (capture, memory, recall, emotion, caregiver) seamlessly. The importance of designing AI systems that are emotionally intelligent, not just functional. How accessibility and UX design greatly impact trust and engagement for elderly users.