SilverArch

Inspiration

Public officers should sleep, but public need does not follow office hours. SilverArch was inspired by citizens who may only be ready to ask for help late at night, especially when facing income loss, caregiving stress, medical bills, or emotional distress.

What it does

SilverArch is an after-hours, voice-first intake assistant for social support services. Citizens can speak naturally in English, Mandarin, Malay, Tamil, Singlish, or dialect, and may mix languages in one call.

The system prepares an officer-ready case packet with:

• timestamped transcript evidence;
• caller rundown;
• missing information;
• safety and urgency flags;
• ComCare and AIC referral considerations;
• callback checklist;
• editable supporting report draft.

SilverArch does not determine eligibility. It helps officers verify faster and respond with better context.

How we built it

We built a local web MVP with a citizen call simulator and an officer dashboard. Audio is recorded or seeded from demo MP3s, transcribed, translated, analysed for key evidence, triaged for scheme/referral relevance, and transformed into an editable report.

Challenges

The hardest parts were handling mixed-language speech honestly, linking every highlight to exact audio timestamps, detecting safeguarding risks, and keeping the officer workflow useful without over-automating judgement.

What we learned

AI is most valuable in public service when it reduces friction while preserving human accountability. For SilverArch, that means: AI prepares; humans decide.

Built With

• css
• elevenlabs-scribe
• express.js
• google-cloud-translation
• html
• javascript
• local-json/file-storage
• mediarecorder-api
• meralion
• node.js
• openai-api
• speechsynthesis-api