SahaAI
A voice-first companion for elderly users, with a dashboard for the people who look after them.
The question
How does an elderly person stay independent while their family and doctor stay informed?
SahaAI gives the elder a phone that talks and listens, and gives caregivers a dashboard for everything the elder has to remember.
Why it exists
Elder care involves the elderly person, their family and their doctor. SahaAI connects the three: a voice-first app for the elder, and a dashboard for caregivers covering prescriptions, reminders, meal plans and alerts.
How it works
How a reminder reaches the elder
- 1
Dashboard
A caregiver adds a prescription or a reminder for the elder.
- 2
API
The backend stores it in PostgreSQL, tied to that elder.
- 3
Phone
The app syncs the reminder into its own local database, so it still works with no connection.
- 4
Alarm
An exact alarm fires and the reminder is spoken aloud. After a reboot the alarms are scheduled again from the local database.
Decisions
Reminders must survive a reboot
A reminder that vanishes when the phone restarts is no use to someone who depends on it. The app uses exact alarms that wake from Doze mode and rebuilds them from the local database at boot.
Two routes to the language model
Chat in the app calls Gemini directly from the phone. Voice chat goes through the backend to Groq, with speech recognition and speech output handled on the device.
Meal plans in a strict format
The meal-plan prompt demands a seven-day plan in a fixed JSON shape, and a second call can refine the plan afterwards.
Evidence
- Built
The backend has about 56 tests covering prescriptions, notifications and the Groq client.
Not built yet
- A family-member portal
- Wearable integration
- Voice in regional Indian languages
Built with
Languages
- Kotlin
Frameworks and libraries
- Jetpack Compose
- Room
- React19
- Vite
- FastAPI
- Prisma
Data and storage
- PostgreSQL
AI and models
- Gemini
- Groq