AWARE

The problem

Most older adults want to age in place, but monitoring fails either the dignity test (cameras, rejected in bedrooms and bathrooms where falls happen) or the reliability test (wearables only work if worn and charged).

How it works

An ~$8 plug-in ESP32-class radio per room reads WiFi Channel State Information — the fingerprint of how WiFi bounces around a space, perturbed by bodies, motion, and breathing. On-device ML turns CSI into presence, room-level location, inactivity flags, learned routine baselines, and respiration-at-rest, with an on-device Gemma model composing caregiver alerts. Inference runs at the edge; only events leave the home.

Key features

• Camera-free, wearable-free CSI sensing on commodity ~$8 hardware
• Presence, room localization, inactivity, routine baselines, respiration-at-rest
• On-device Gemma 4 E2B reasoning (7 modes) with a rule-based fallback
• Caregiver + community-responder alerting from a single dual-role Android APK
• Honest capability tiers (ship-now vs. emerging vs. roadmap) — explicitly not a medical device

Architecture

2× ESP32 stream CSI over UDP to a host Android phone → breathing FFT DSP + state aggregator → Gemma 4 E2B every 30s → alerts published to a caregiver phone. A cloud pitch-tier path (Pub/Sub → Cloud Run → Firestore → FCM) scales it to a pilot.