CuraLink

Helping people with Alzheimer’s stay independent and safe, with caregivers always connected.

Comment

Janhavi Dhote, sewon Myung

Inspiration

Alzheimer’s is not just about forgetting names.

It is about losing orientation.
To space.
To people.
To routine.
To emotional safety.

We were inspired by how often people with Alzheimer’s feel confused or anxious in everyday moments — especially when no caregiver is nearby.

Most tools focus on tracking or alerts.
We wanted to focus on support in the moment.

Something calm.
Something respectful.
Something human.

What it does

CuraLink is an AI assistant designed to support people with Alzheimer’s and their caregivers.

It helps users:

  • Understand what they are looking at
  • Recognize familiar people
  • Stay oriented to place and time
  • Remember daily routines
  • Feel calmer during moments of anxiety

It also supports caregivers by:

  • Allowing them to register faces and routines
  • Monitoring how the system responds
  • Staying in control of what the AI learns

CuraLink does not replace caregivers.
It supports them — while helping patients stay more independent.

How we built it

CuraLink is built primarily in Python.

We used FastAPI and Uvicorn to build a lightweight backend, with a simple frontend using HTML and Jinja2 templates.

For computer vision and physical awareness, we used:

  • YOLOv8 for real-time object detection
  • OpenCV and NumPy for image processing
  • InsightFace with ONNX Runtime for face recognition and embeddings

For voice and conversation, we used:

  • SpeechRecognition and PyAudio for listening
  • OpenAI’s GPT API for reasoning and dialogue
  • Text-to-Speech using OpenAI, with optional local support (Coqui or Piper)

For navigation and orientation:

  • OpenRouteService API for turn-by-turn directions

For data storage:

  • Local JSON files for routines, locations, and voice recordings
  • NumPy (.npz) files for face embeddings
  • A file-based local database, with future plans for PostgreSQL and vector search

The system includes:

  • A patient-facing interface designed to be calm and minimal
  • A caregiver dashboard where all learning is reviewed and approved

Challenges we ran into

Designing for Alzheimer’s care is hard.

Too much information can overwhelm users.
Too little information can increase anxiety.

We had to be careful with:

  • Face recognition consent and accuracy
  • Routine creation timing
  • Emotional responses during stress or confusion

Another challenge was shifting our mindset.
This was not about building the smartest AI.
It was about building the kindest one.

Accomplishments that we're proud of

  • Building a visually aware AI that goes beyond object detection
  • Creating a face recognition system that learns over time with caregiver approval
  • Designing a calm, non-alarming voice experience
  • Integrating emotional support through familiar voices
  • Keeping caregivers in control at every step

Most importantly, we built something that feels human.

What we learned

We learned that accuracy is not everything.

For Alzheimer’s care:

  • Reassurance matters more than precision
  • Familiarity matters more than speed
  • Calm matters more than complexity

We also learned the importance of human-in-the-loop AI. Caregivers must always have control.

Technology should adapt to people — not the other way around.

What's next for CuraLink

Next, we want to:

  • Improve stress and anxiety detection
  • Expand routine learning using visual context
  • Add secure cloud support for caregivers
  • Improve personalization for different stages of Alzheimer’s
  • Continue working with real caregivers for feedback

CuraLink is not finished.

Built With

Share this project:

Updates