Lumos

Architecture Diagram
Dain Agent
Dain Agent
Spectacle's View

Inspiration

Watching loved ones struggle with Alzheimer's made us realize how painful it can be to lose not just memories, but a sense of connection. We were inspired to create a system that could gently assist memory recall without being invasive — offering a bridge between technology and human dignity.

What We Built

LumosCare is a comprehensive memory assistance system designed to support both patients with Alzheimer's and their caregivers through a seamless and dignified experience. It includes:

1.⁠ ⁠AR Glasses Interface : Patients wear Snapchat Spectacles that recognize familiar faces and display memory cues like names and relationships using subtle AR overlays.

2.⁠ ⁠AI Agent System : We developed autonomous agents using Fetch.ai, Dain and Gemini 2.5 to handle face recognition, contextual reminders, safety monitoring, and daily journaling prompts.

3.⁠ ⁠Secure Backend Infrastructure : A backend built with Node.js, MongoDB Atlas, and oAuth securely stores patient profiles, caregiver data, safe zones, and memory logs. Used Melissa API to find the nearest hospitals for the patients.

4.⁠ ⁠Caregiver Mobile App : A React Native app empowers caregivers to manage profiles, set reminders, create safe zones, and receive real-time alerts about the patient’s activity and safety.

What We Learned

-- ⁠How to orchestrate multiple AI agents to create a natural, seamless support system.

--⁠ ⁠How to optimize AR overlays for real-time use on hardware with limited resources.

-- ⁠How crucial it is to design systems that prioritize user privacy and emotional well-being.

Challenges We Faced

Building LumosCare wasn’t without its obstacles:

1.⁠ ⁠Reliable face recognition on AR glasses — Achieving low-latency, real-time recognition on hardware with limited processing power required extensive optimization and testing.

2.⁠ ⁠Seamless multi-service integration — Combining real-time geofencing, face recognition, and voice journaling into one smooth experience took a lot of trial, error, and careful orchestration.

3.⁠ ⁠Designing for ethical use — We had to think deeply about privacy, ensuring that the technology empowered users without ever feeling invasive or controlling.