Memorie

Novel PWA powered by AI & compatible with XR devices that passively detects memory decline + provides memory assistance to Alzheimer's patients, helping them recognize loved ones & remember things.

Comment 3

Remember the faces that matter.

💡 Inspiration

Watching loved ones struggle with Alzheimer’s and other forms of dementia 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 passively detect memory decline and— for those who do develop Alzheimer’s—gently assist memory recall without being invasive, turning moments of uncertainty into moments of renewed connection.

🌍 Potential Impact

  1. 57 million people worldwide are living with some form of dementia in 2025.
  2. The World Health Organization projects this number will triple to 152 million by 2050.
  3. 50 % of cases go undiagnosed. Earlier detection unlocks treatments that slow progression and could save the world trillions in cumulative medical and long-term-care costs.

👀 What We Built

Imagine a doctor telling you that you’ve been diagnosed with Dementia or Alzheimer's—but because you were tested early, you can slow down its decline by 18 months or more. And that even after memory starts to fade, you can still remember important people and details, so you can maintain independence and confidence in daily life.

Introducing Memorie—a novel comprehensive memory-assistance app that passively detects memory decline (early detection) and supports both patients and caregivers through a seamless, accessible experience.

  1. World’s first Progressive Web App (PWA) compatible with XR devices 100 % built and run on Bolt.new.
    Meta recently rolled out experimental access to their passthrough camera API, just allowing this to be possible. Running it as a PWA gives access to anyone in the world with a camera-supported device (mobile, tablet, etc) and internet connection, which is great for people who don’t have access to AR glasses.
  2. World’s first Passive Early Detection for Dementia via Facial Recognition
    The leading causes for going undiagnosed are a lack of healthcare access and stigma. Our app solves both of these problems. Our Memory Score algorithm runs in the background, eventually with no user input, to track memory over time—reducing stigma and flagging when clinical testing (MMSE, MoCA, CDR, etc.) might be warranted.
  3. AR-Compatible Interface
    Patients wear any AR-supported devices/glasses that recognize familiar faces and display memory cues like names, relationships, and notes using their browser as an AR overlay.
  4. Caregiver Profile
    Our app empowers caregivers to manage profiles, add reminders, and see real-time patients’ activity of those they interact with.

✅ How to Use It

In three simple steps, Memorie tracks your memory score and displays a person’s profile next to their face on any camera-enabled device with internet:

  1. Add a contact
  2. Take a picture of the familiar face
  3. Select whether you recognized the person (optional for Memory Score)

⚒️ How We Built It

100 % of the code is generated and runs within Bolt.new.

PWA App

  • Front-end: React.js
  • Back-end: Node.js
  • Database: DynamoDB
  • User photos: AWS S3
  • AI facial recognition: AWS Rekognition

🎓 What We Learned

  1. Building PWAs powered by AI that are XR-compatible.
  2. Optimizing web apps as AR overlays for real-time use on hardware with limited resources.
  3. Designing systems that prioritize emotional well-being for vulnerable users.

🏃‍♂️ Challenges We Ran Into

  1. Camera access on Meta Quest: Figuring out how to build and run an AR app that has camera access 100% within Bolt was difficult. Went with a PWA so it’s supported across all mobile devices (iOS & Android) and all future AR glasses that run on Android XR (Meta's Hypernova, Google's XR glasses, Samsung's Moohan and Haean, Xreals Aura, etc).
  2. Multi-platform sync: Ensuring real-time consistency across three frontend interfaces (web, mobile, AR) was complex.
  3. Emotion-centric design: We had to think deeply about memory tracking, ensuring that the technology empowered users without it feeling self-critical or invasive.

🚀 What’s Next

  • Multimodal Memory Score: Voice analysis for frictionless, more accurate tracking.
  • Multilingual support for a broader, global demographic.
  • Auto-import contacts and photos from phones and cloud address books.
Share this project:

Updates