For families & memory-care teams
Bring them back to the songs they still know by heart.
Familiar is an AI agent that builds a personalised Memory Kit for someone living with dementia: the music, films, food and stories of their youth, from their hometown and culture. It is grounded in Qloo's taste graph of 250M+ cultural entities and learns from every session.
Session · step 3 of 6
Play something she danced to
Put on a bolero from the 1960s, hold her hands and sway gently.
Say
"Your sisters told me you were the best dancer at every wedding. What did you wear?"
Every reaction becomes a signal for the next version of the kit.
57M
people live with dementia worldwide, with ~10M new cases every year (WHO).
Ages 10–30
Memories from these years survive longest. It's called the reminiscence bump.
Music
Musical memory is often among the last to fade, and personalised playlists are used to ease agitation.
The problem isn't music. It's which music.
What a care home usually plays
One "golden oldies" playlist for every resident born in the 1940s. A generic chatbot does the same: it guesses from stereotypes about an age group.
What Familiar does
It crosses era × hometown × heritage × the things they already love in Qloo's taste graph. A Tejana grandmother from San Antonio and a Motown fan from Detroit, both born in the 1940s, get completely different kits, and every item is a real, verified entity.
How it works
- 1
Tell their story
Five minutes: birth year, where they grew up, their culture, a few things they've always loved, and anything to avoid.
- 2
The agent researches
Familiar verifies their favourites in Qloo, finds anchors from their formative years, and expands outward through the taste graph by age, place and affinity.
- 3
Run a session and it learns
A calm, large-type session guide. Tap how they responded. What lit them up becomes the next signal; what upset them is excluded for good.
Three people, three different worlds
Fictional people born within eleven years of each other. See the kits Familiar built for them, with every research step, or run the agent live yourself (about a minute).