Relief - your virtual presence for loved ones with Alzheimer's
Team led by an AI/ML CTO (Python, TypeScript, ex-InVision) and a Product/UX Co-founder building agentic video creation tools.
YouTube Video
Project Description
For over 10 years, I’ve watched my grandmother’s Alzheimer’s progress and my mother devote herself completely to caring for her. Supporting a loved one with Alzheimer’s is both emotionally and physically demanding. As memory and reasoning fade, simple conversations and repetitive questions can turn into moments of confusion or frustration for both the individual and the caregiver. On top of this, caregivers often feel they must be constantly present. Over time, that unrelenting pressure can lead to deep stress, exhaustion, and burnout.
As assistive technology continues to evolve—and with emerging research on conversational agents designed to support people with dementia and their caregivers—we saw an opportunity. We imagined an AI tool that could help ease both sides of the experience: creating calmer, more patient interactions for the person with Alzheimer’s, while giving caregivers space to step away when needed without guilt.
Relief is a multimodal Virtual Presence Agent that uses advanced voice and vision capabilities to offer real-time, familiar-feeling conversations for people with Alzheimer’s. When a primary caregiver can’t be present, Relief provides comforting, supportive interactions—offering reassurance for loved ones and meaningful peace of mind for caregivers.
The working prototype was built for demonstration purposes. The application is a thin front-end layer, using n8n webhook for analizing the webcam’s images and sending it to the ElevenLabs agent as extra context. It also pushes extra context for keeping up the conversation to avoid idle time.
With more time, we’d add
- Creating a summary for the caretaker
- Logging interaction data to track the progress of the dementia
- Notify the caretaker in case of emergency
We used react, tailwind, n8n, ElevenLabs agent, GPT-5-mini to create this proof of concept.
The app works both on mobile devices and in desktop browsers.