A safety-moderated carevatar platform that extends family presence for people living with dementia, without pretending to be family. Built for the caregiver first.
The hard moments rarely arrive during a planned visit. They arrive at two in the morning, mid-agitation, or in the fortieth repetition of a question that deserves as much patience as the first.
A carevatar is a caregiver's instrument for those hours: a familiar presence delivered through a simple video-call interface, speaking from memories and routines the caregiver has curated and approved.
It extends the family's presence. It does not replace it, and it does not pretend to.The single design refusal: warmth without impersonation.
A tool for vulnerable people should announce its own limits more loudly than its features.
Dementia care cannot ride on a single model and a content filter. Creation, moderation and evaluation are separated so no single failure reaches the person on the screen.
Speaks only from caregiver-approved memories, scripts and routines, inside a session the caregiver started.
Screens every exchange before and after generation, and can redirect, refuse or end a session.
Watches patterns over time and surfaces observations for the caregiver and physician, never a diagnosis.
The dementia-companion market is full of engagement statistics and empty of clinical constraint. We are taking the slower road: an eight-week feasibility study, CARE-SAT, with twenty caregiver–patient dyads, pre-specified endpoints, and a definition of failure written down before the first session runs. Satisfaction, caregiver burden, mood, sleep, and agitation are measured with standard instruments; safety events are defined, logged, and reported whether or not the results flatter us.
The ethical framework we build against is published independently by the Institute for AI Ethics in Human Cognition, whose oversight standard, and whose insistence that responsibility stays with humans, precedes and outranks this product.
The architecture is specified, the safety policies are written, and the test scenarios exist as structured cases. What the project needs now is a small number of serious people.
Juan Pablo Buritica leads technical development of the MVP, including backend implementation, model routing, safety evaluation, and the behavioral test infrastructure.
Read the developer brief →Founding, unpaid, flexible. We need clinicians, bioethicists, and HCI researchers who know this terrain to tell us where the design is wrong, before the pilot runs, not after.
See what advising looks like →Lucid Bridge is currently seeking external funding to support MVP development, hosting, CARE-SAT preparation, participant coordination, and independent evaluation.
Read the prospectus →We are looking for caregivers, clinicians, researchers and funders who care more about constraint than hype.Brian Harris, MD · Founder and Medical Director, EusomniaMD
Or simply write: bharris@eusomniamd.com. The 14-slide Carevatar pitch deck is the fast tour.