SMRITI

A local-first household system for an older adult and their family. It runs on the home Wi‑Fi, and nothing goes to the cloud.

Status: Prototype
Year: 2026

Briefing

Problem: Continuity at home for an older adult, without handing their life to a vendor cloud.
What exists today: A prototype that runs on the home Wi‑Fi, with an optional play-pace model trained on synthetic data only. Not a medical device.

Objective

Continuity at home for an older adult, without handing their life to a vendor cloud. SMRITI is not a medical device: it does not diagnose anything, medicine taps are self-reported only, and the older adult never sees a score.

Architecture

Schematic

Companion app
- Large type, voice-friendly
- Used by the older adult
Link: Home Wi‑Fi
Home server
- Runs on a laptop
- Express, SQLite, Socket.IO
Link: Home Wi‑Fi
Family Care app
- Pairing, reminders, memories
- Used by the family

Both phones talk to the laptop over the home Wi‑Fi. Pairing uses a short code or a QR code, so the older adult never needs an email address.

Constraints

Every record says who reported it and when. When two reports disagree (for example, a relative saw the tablets still in the box but the companion phone logged them as taken), the app flags it for review instead of choosing one.

Measurements

After a play session, an optional local model suggests a gentler or fuller game pace. It came out of a Smart India Hackathon 2026 problem on cognitive gaming and memory support for older adults with dementia in India’s North Eastern Region. It was trained and tested on synthetic play events only: a research prototype, not a diagnosis.

LSTM training curves from the SMRITI play-pace model, trained on synthetic play events. — Training curves, synthetic data.

Confusion matrix of the SMRITI play-pace model on its synthetic test split, across three non-clinical concern levels. — Test-split confusion matrix, synthetic data. Categories are non-clinical.

Status

Prototype