Abstract
As Singapore matures into a Smart Nation, a silent "language divide" persists between the dialect speaking elderly and their foreign caregivers. This communication gap often leads to medical errors, misunderstanding of daily needs, and deepening social isolation. We propose Silver Token, a "Zero-UI" translation token powered by MERaLiON, Singapore’s sovereign AI model. Unlike generic apps that fail on local nuances, our device accurately interprets code-switched Singlish (e.g. “Makan”, “Shiok” and “Alamak”) and diverse dialects. The elderly simply "push-to-speak," eliminating the cognitive burden of complex touchscreens. Crucially, we integrate a "Digital Leash" for robust security. Silven Token utilises dynamic network zoning to authenticate safety. It only functions when tethered to trusted "Safe Zones" such as Home Wi-Fi or Caregiver Hotspots. If the device leaves this trusted proximity, it automatically locks, preventing unauthorized data access. Furthermore, we guarantee Data Sovereignty by hosting all voice processing within Singapore, ensuring full PDPA compliance. Built on Green AI principles for extended battery life, Silver Token transforms technology from a barrier into a lifeline. By bridging linguistic gaps securely, we empower our pioneers with independence, ensuring they are connected, protected, and never left behind.
Inspiration
Silver Token was inspired by my grandmother. At her age, something as simple as ordering pizza from Domino’s or Pizza Hut becomes stressful because she can’t communicate clearly in English. Polyclinic appointments give her even more anxiety—she worries the doctor or nurse won’t really understand her symptoms. We realised she’s not alone; many dialect-speaking seniors quietly struggle with the same fear of “saying the wrong thing” every day. Silver Token is our attempt to give them back confidence and independence.
What it does
Silver Token is a zero-UI translation token for dialect-speaking seniors and their foreign caregivers. The elderly just push-to-speak in the language they’re comfortable with—Hokkien, Teochew, Malay, code-switched Singlish—and the token speaks out a clear translation for the caregiver, and vice versa. No apps, no touchscreen, no updates to manage. We also designed a “Digital Leash” concept: the token only works in trusted safe zones (like home Wi-Fi or a caregiver hotspot), and all voice processing is intended to stay within Singapore for data sovereignty and PDPA compliance.
How we built it
We started by mapping real scenarios from my grandmother’s life—ordering food, describing pain, talking to clinic staff—and turned them into user stories. On the tech side, we prototyped a speech pipeline in Python: recording audio, sending it to an AI model for transcription and translation, and playing back natural speech. This first version uses cloud AI (e.g. Gemini) but is designed to be swapped to MERaLiON, Singapore’s sovereign AI model. In parallel, we designed the hardware stack around a Raspberry Pi Zero 2 W, a microphone HAT, a large arcade button, and a simple LED feedback system to keep the interaction physical and intuitive.
Challenges we ran into
Hosting AI locally was a big challenge—running large speech and translation models on our own machines is slow and resource-heavy, so we had to balance between edge processing and cloud calls. Another challenge was designing from an elderly perspective. As 20-something developers, it’s very easy to assume comfort with screens, text, and notifications; forcing ourselves to strip everything down to “one button, one action” was surprisingly hard. Finally, dealing with dialects and messy Singlish (like “alamak”, “makan”, “shiok”) showed us how fragile generic translation APIs can be.
Accomplishments that we're proud of
We’re proud that we moved beyond just a slide deck into a working end-to-end demo: from live audio input to AI translation to spoken output. We reduced the interaction to a single, elderly-friendly gesture while still keeping space for security features like the Digital Leash. As a team, this was our first time combining low-power hardware, AI speech services, and a social problem in one prototype, and seeing it actually translate real Singlish phrases was a huge milestone.
What we learned
We learned that AI for seniors is not just about accuracy—it’s about accessibility. Even the best translation app is useless if the user can’t unlock the phone, find the icon, or read tiny fonts. The smartphone itself is a massive friction point: updates, passwords, random pop-ups. That’s why we shifted hard towards a token form factor with no screen and no menus. On the tech side, we also learned a lot about latency, model limitations with dialects, and how important it is to use local data and local models for Singapore’s unique mix of languages.
What's next for Silver Token
Next, we want to finish the physical prototype and bring it into the real world. Our plan is to test Silver Token with seniors at community centres and elderly homes, alongside social workers and caregivers, to collect feedback on comfort, trust and real-life usefulness. Based on those insights, we’ll refine the hardware, tweak the prompts and flows, and explore integrating MERaLiON for fully local, sovereign AI. From there, we hope to partner with agencies and clinics to turn Silver Token from a hackathon project into something that genuinely helps our pioneers feel understood.
Built With
- meralion
- python
Log in or sign up for Devpost to join the conversation.