BridgeCare

Inspiration

BridgeCare was inspired by someone very close to me—my grandmother.

She lives at home with a caretaker, while both of her children—my mom and my uncle—live abroad. They love her deeply, and they talk to her every day, but because of work, distance, and time zones, it’s simply not possible to be on a call with her 24/7. That leaves many quiet moments in the day where she feels lonely.

When I was younger, I grew up with her. She was the heart of the home—active, energetic, and always taking care of everyone. During COVID, my grandfather passed away, and my grandmother also became very ill. She eventually recovered, but she is not able to manage the house the way she used to, and now she needs a caretaker.

What affected me the most was not the physical change, but the emotional one. Seeing her feel lonely was very difficult, because she is one of the most loving people in my life.

BridgeCare is not meant to replace family. No technology can replace real human connection. Her children still call her every day, and that matters the most. But I kept thinking: what if there was something gentle and supportive that could talk to her during the quiet moments in between?

If an AI companion could reduce even a small amount of that loneliness, it would make a meaningful difference. That idea became BridgeCare.

What it does

BridgeCare is an AI-powered voice companion designed for elderly users.

Instead of typing into an app, the user simply presses a button and speaks naturally. The system listens, responds with a warm and empathetic voice, and quietly analyzes emotional signals in the conversation.

Each interaction produces:

An empathetic AI reply
An Emotional State Indicator (ESI) score from 0–100
A simple risk level
A short, family-friendly explanation of the user’s emotional state

This allows families to stay emotionally connected without reading full transcripts or invading the user’s privacy.

BridgeCare also uses personal memories—such as family routines or preferences—to make responses feel more familiar and human.

How we built it

We built BridgeCare as a voice-first, AI-powered system using a lightweight architecture suitable for a hackathon prototype.

Frontend

React with TypeScript
Tailwind CSS for the interface
Web Speech API for speech-to-text and text-to-speech

The user presses “Start,” speaks naturally, and the transcript appears in real time. The AI’s response is displayed and spoken back to the user.

Backend

Node.js with Express
Google Gemini API for empathetic conversational responses
A lightweight memory retrieval system (RAG-lite)

The backend:

Receives the user’s speech transcript.
Retrieves relevant personal memories.
Sends both to the AI model.
Returns a structured response including the reply, emotional score, risk level, and explanation for family.

This approach allowed us to create personalized, emotionally aware conversations without heavy infrastructure.

Challenges we ran into

Voice reliability Browser-based speech recognition can be unpredictable. We had to handle microphone permissions, random stoppages, and fallback interaction modes to keep the demo stable.

Balancing empathy and safety We wanted the AI to feel warm and human, while ensuring it did not give medical advice or sound overly clinical.

Personalization under time constraints We wanted memory-based responses, but we had limited time. We built a lightweight retrieval system that simulates long-term memory without complex infrastructure.

Frontend–backend integration We developed the voice interface and the AI backend separately, then integrated them near the end. Debugging runtime errors and speech issues under time pressure was one of the biggest challenges.

Accomplishments that we're proud of

Built a fully working voice-based AI companion in a hackathon timeframe
Created an emotional scoring system that translates conversations into simple insights
Implemented memory-grounded responses to make interactions feel more personal
Designed a system that respects privacy and does not replace human relationships
Turned a deeply personal inspiration into a functional, real-world prototype

What we learned

Voice interfaces require careful UX design to feel natural and reliable.
Emotional AI must be simple, gentle, and non-intrusive.
Personal context and memory dramatically improve user connection.
In hackathons, demo stability and clarity matter more than perfect architecture.
The most meaningful projects often come from real human problems, not just technical ideas.

What’s next for BridgeCare

In the future, we would like to:

Add multilingual support so seniors can speak in their native language
Build a dedicated caregiver dashboard with emotional trends over time
Introduce daily check-ins and gentle reminders for routines or medications
Integrate with wearable or smart home devices for a more complete well-being picture
Deploy the system on a tablet or smart speaker for easier access

Our long-term vision is to create a simple, ethical, and emotionally supportive AI system that helps seniors feel heard—and helps families stay connected, no matter the distance.