Revere

• 

• 

• 

• 

Inspiration

Alzheimer’s disease often affects a person’s ability to recognize familiar individuals and retain important day-to-day information, creating stress for both patients and caregivers. We were inspired to build Revere as a practical assistive system that uses AI to support memory, reduce confusion, and improve quality of life.

What it does

Revere is an AI-powered wearable and caregiver platform designed to support individuals with Alzheimer’s in real time. The system recognizes familiar people, delivers contextual audio cues, and provides caregivers with a centralized dashboard to manage reminders, monitor interactions, and retrieve system information through a voice-enabled assistant.

Feature	Implementation	Impact
Face Recognition	Computer vision on Raspberry Pi wearable	Helps users identify familiar people
Audio Cues	Real-time spoken prompts based on recognition	Reduces confusion and supports memory
Caregiver Dashboard	Next.js dashboard with Firebase backend	Centralized caregiver controls and monitoring
Voice Assistant	ElevenLabs STT + TTS integration	Accessible hands-free caregiver interaction
Cloud Sync	Firebase + Vercel Blob for storage and sync	Reliable data access across the system

How we built it

We developed Revere as a full-stack solution that combines a Raspberry Pi-based wearable device with a web-based caregiver dashboard. The wearable uses computer vision for face recognition, while the dashboard is built with Next.js and connected to Firebase for primary data storage and management and Vercel blob for image storage. We also integrated ElevenLabs speech-to-text and text-to-speech APIs to create an accessible voice interface for caregivers.

Challenges we ran into

API/ Database costs: We tried using a pro version of gemini for better spatial detection but incurring token charges surmounted, so we had to stay within free limit and use Gemini Flash 2.0.

Real-time sync - Buffering between 2s frontend polls and faster backend processing, polling api calls from raspberry pi zero 2w

GSAP-WebGL scroll scrubbing - Making 3d model animations snappy and lag-free was a tiring goal we are proud of successfully achieving

Accomplishments We're Proud Of

We successfully synchronized hardware, computer vision, and cloud infrastructure into a single cohesive ecosystem. Instead of separate "duct-taped" demos, we built:

Wearable Input: Real-time face recognition.

Live Feedback: Data feeding into a caregiver dashboard.

Accessibility: A voice-enabled layer for the system.

We built a front-end that balances tech with human empathy:

Visuals & Motion: Made using Higgsfield and Gemini Veo 2.0.

Assets: 3D elements via Meshy.AI and Canva identity.

What We Learned

Bridging the Gap: Integrating hardware with web stacks.

Human-Centric Design: Tech must be dependable and sensitive.

AI Orchestration: Using AI for efficiency and aesthetics.

What's Next for Revere

Performance: Porting to a high-performance CPU.

Intelligence: Expanding contextual AI reminders.

Hardware: Refining the wearable form factor.

Interaction: Enhancing the voice assistant dialogue.

Built With

• c
• firebase
• geminiapi
• next.js
• python
• raspberry-pi
• typescript