NeuroBridge

Inspiration

25% of autistic people are non-verbal or minimally verbal. 65% live with chronic pain they cannot describe. When a person cannot say "I'm hurting" or "I'm scared," caregivers are left guessing. And guessing isn't caring — it's a health crisis.

NeuroBridge was built to close that gap. We wanted to give every non-verbal individual a voice, and every caregiver a clear signal. Not a diagnosis. Just a bridge.

What it does

NeuroBridge is an AI-powered communication aid that turns pictogram taps into natural, spoken sentences — in under one second.

A user selects one or several icons: "hurt," "too loud," "anxious," "hungry," "sleepy," or "I need help." NeuroBridge's personal AI agent interprets the combination and generates a clear first-person sentence, such as:

"I feel overwhelmed because it is too loud. I need a quieter place."

The app reads the message aloud using text-to-speech, so the user can communicate without needing to speak.

For caregivers, NeuroBridge includes a dashboard that stores every exchange with a timestamp, highlights recurring patterns, and allows caregivers to better understand the person's evolving needs.

How we built it

NeuroBridge is a web application built with Next.js, React, TypeScript, and SQLite. The interface is designed around 24 ARASAAC pictograms — the international standard used by speech therapists — grouped into categories: needs, pain, emotions, and sensory experiences.

The AI layer transforms selected pictograms into natural language using an OpenAI-compatible API. Each user has a personal AI agent that learns their patterns over time, making interpretations more accurate with repeated use. We used the browser's built-in speech synthesis API to read messages aloud.

The caregiver dashboard was built with a focus on clarity: timestamped history, pattern detection, and an AI assistant that can answer questions based on the user's communication log.

Challenges we faced

One major challenge was making the AI responses helpful without overcomplicating them. The generated sentences needed to sound human, warm, and emotionally clear, while remaining short enough for real-time communication. We iterated extensively on prompt engineering to find the right balance.

Another challenge was designing the experience to support neurodiverse users without pretending to be a medical diagnosis tool. NeuroBridge is not a medical device — it is a communication and accessibility aid. We made this clear throughout the product and deliberately avoided any diagnostic or clinical claims.

We also faced the challenge of making the interface accessible: 56px touch targets, 7:1 color contrast, and full screen-reader support were non-negotiable from day one.

Accomplishments

We are proud that NeuroBridge combines accessibility, AI, and caregiver support in a working prototype that can be tested directly in the browser — with no account required.

Key features we delivered:

• 24 ARASAAC pictograms for intuitive communication
• AI-generated first-person sentences from pictogram combinations
• Voice synthesis for instant spoken output
• Personal AI agents that learn individual patterns over time
• User profiles and timestamped communication history
• Caregiver dashboard with pattern recognition
• Privacy-first design: no data sharing, no ads, no mandatory accounts
• Accessibility-first interface: 56px targets, 7:1 contrast, screen-reader support

What we learned

We learned that AI can be useful not only for productivity or automation, but also — and perhaps more importantly — for making communication more inclusive and human. A small interface decision, like the size of a touch target or the tone of a generated sentence, can have a profound impact when the user has limited verbal communication.

We also learned the importance of building responsibly when dealing with sensitive personal experiences, especially around disability, caregiving, and communication data. Privacy isn't a feature — it's a foundation.

Finally, we learned that the best technology is invisible. When NeuroBridge works, the user doesn't think about AI or APIs. They just communicate.

What's next

We have a clear roadmap for NeuroBridge:

• Stronger privacy & authentication — encrypted profiles and optional secure login
• More pictograms & customization — expanded library and personalized vocabulary
• Multilingual support — reaching non-English speaking communities worldwide
• Offline-friendly mode — communication without internet dependency
• Better caregiver insights — advanced pattern analysis and trend detection
• Safer AI fallback responses — graceful degradation when AI is unavailable
• Real-world accessibility testing — with autistic individuals, caregivers, and speech therapists

Built With

• better-sqlite3
• build-with-afri
• css
• html
• javascript
• next.js
• openai-compatible-api
• react
• speech
• sqlite
• typescript
• web