CASE 001 / AI / DEMENTIA & ELDERLY CAREPROTOTYPE
WE care
Voice-first, caregiver-supervised AI companion for dementia and elderly care.
01 · Problem & Constraints
The team's problem statement — revealed on-site at DRESTEIN '26 — was to design a voice-first, stage-adaptive AI companion for dementia patients that caregivers could actually trust and supervise.
- AI must not invent facts about the patient — memory has to be caregiver-verified
- Caregivers need real-time visibility and the ability to escalate
- Consent and access control across patient/caregiver roles
02 · Architecture & Flow
Explicit Perimeters & Component Flow
01Stage-adaptive voice companion (Android, Jetpack Compose, MVVM, Hilt)
02Distress risk scoring (Green / Amber / Red) with human-in-the-loop escalation
03Real-time caregiver alerts via Supabase Realtime and Firebase Cloud Messaging
04Next.js + React web dashboard for caregivers
Technology Stack
KotlinJetpack ComposeMVVMHiltNext.jsReactSupabaseFirebase
03 · Security Decisions
Trust Boundaries Made Legible
- Caregiver-verified memory prevents AI-invented facts
- Explicit consent workflows
- Role-based access via PostgreSQL row-level security (RLS)
04 · Outcome & Roadmap
Verified Results & Real-World Testing
1st Prize, ₹1,00,000 — DRESTEIN '26 (Saveetha Engineering College). Took the idea from R&D to a working Android and web prototype in two days.
Next Iterations
The team intends to keep developing it toward a production-ready product rather than stopping at the hackathon prototype.