PicoCare
Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

PicoCare is a personal care secretary for older adults living alone, with a companion Android app for family caregivers and care teams. It addresses a familiar worry: when a parent goes quiet, families lack context, while fixed check-in alarms can interrupt ordinary life. With the person's chosen permissions, PicoCare collects everyday phone clues locally: screen activity, steps, media playback, navigation categories, movement summaries, and battery changes. Meta Muse Spark 1.3 combines these clues into concise, evidence-backed observations. When something needs clarification, PicoCare asks the older adult first, using simple answer buttons and a question-specific "None of the above" response. Explicit help requests go directly into the paired caregiver's help workflow; unresolved concerns include the evidence and the person's original reply. Two core scenarios are deferring a check-in during sustained video playback and asking a respectful question when outdoor movement and navigation patterns change. A timestamped Secretary Feed keeps the reasoning visible. A weekly family reflection uses Meta's open-weight Muse Spark model through Baseten, with the model source shown in the UI. Built during the hackathon with Kotlin Android apps, a TypeScript/Node.js backend, SQLite, and a caregiver web console. English and Traditional Chinese are supported. Local evidence storage and an offline outbox preserve observations and answers through connection loss. The demo includes real Samsung Galaxy S23 screens. PicoCare supports everyday human follow-up; it provides no medical diagnosis or advice. Care should feel like care. The older adult gets the first word.