Inspiration

We built this because every time we talked to an elderly person, we realised how often they say, “Aiya, I don’t know how to use all these apps lah.” Not because they don’t want to join activities or stay connected — but because technology isn’t built for the way they communicate.

Most seniors are more comfortable talking than typing, often mixing Singlish, pauses, dialect, and emotion. Existing systems don’t handle that. SilverCompanionSG began with one question:

“What if seniors could talk the way they normally do — and the system understood them?”


What it does

SilverCompanionSG is a voice-first companion that lets seniors:

  • Find and register for community events
  • Trigger emergency alerts
  • Have casual conversation
    All through natural speech, without needing to navigate complex apps.

How we built it

  • Whisper for speech-to-text
  • OpenAI for translation and intent detection
  • FastAPI orchestrates actions
  • Supabase stores events and user info
  • Twilio triggers SOS alerts

Workflow: user speaks → Whisper → OpenAI interprets → FastAPI executes → Supabase/Twilio respond.


Challenges we ran into

  • Whisper mishearing Singlish
  • Seniors phrasing requests indirectly
  • Voice currently supports only a subset of features; others still need UI navigation

Accomplishments that we're proud of

  • Fully functional voice-first workflow for seniors
  • SOS alerts working in real-time
  • Seamless integration between STT, GPT, and backend

What we learned

Designing for seniors requires empathy with natural speech, indirect phrasing, and error-handling in real-world voice systems.


What's next for SilverCompanionSG

Expand voice support to all features, add context awareness, and improve offline/fallback functionality.

Built With

Share this project:

Updates