Inspiration

NeuroDrive Helper was inspired by the need to make online reading more accessible for individuals with ADHD, dyslexia, and other neurodiverse conditions. Drawing from multiple scientific reports, we were able to compile clear definitions of numerous conditions and their corresponding treatments. We even related to one of our teammate's cases, where he deals with red-green colorblindness, which inspired us to add features that assist with others who deal with those same issues. Many websites present dense, visually cluttered text that can be overwhelming to process. Our goal was to create an AI-driven tool that adapts web content to different cognitive needs—simplifying language, improving focus, and offering flexible reading options.

What It Does

NeuroDrive Helper is an AI-powered Chrome extension that simplifies, summarizes, and reads aloud any webpage. It rewrites complex text into clear, concise language, highlights key ideas, and provides accessibility features such as dyslexia-friendly fonts, adjustable spacing, high-contrast mode, and a distraction-free reading overlay and the type of color-blindness. Users can also activate text-to-speech to listen to content in real time.

How We Built It

The project combines a Chrome Extension frontend with a FastAPI backend. The backend connects to Hugging Face language models for summarization and simplification, and to ElevenLabs for text-to-speech synthesis.

Frontend: JavaScript-based Chrome extension with a popup interface, content scripts, and a service worker for background processing.

Backend: FastAPI in Python using models such as google/flan-t5-base and sshleifer/distilbart-cnn-12-6.

Infrastructure: Dockerized for consistent deployment and CORS-enabled for browser communication. Text is sent to the backend, processed through the model pipeline, and returned as structured, simplified content for on-page display.

What We Learned

We learned how to integrate large language models via the Hugging Face API, manage asynchronous messaging between Chrome extension components, and design for accessibility following WCAG guidelines. We also gained experience in containerizing applications with Docker and handling model errors gracefully using fallbacks and retries.

Challenges

We faced challenges with API latency, managing model errors such as 404 and 503 responses, and ensuring the extension’s injected UI did not interfere with webpage layouts. Handling long text through chunking and maintaining readability across chunks also required careful tuning.

Future Work

Future improvements include personalization based on user reading profiles, local inference for offline use, and expanded browser support. We also aim to add adaptive reading modes and further refine accessibility customization.

Conclusion

NeuroDrive Helper demonstrates how AI and accessibility engineering can combine to create inclusive digital experiences. This project reflects an effort to make the web clearer, calmer, and more readable for everyone.

Built With

• chrome
• chrome-extension-apis-(frontend)-ai-&-nlp:-hugging-face-inference-api-(google/flan-t5-base
• chrome-storage-api
• css
• css-frameworks:-fastapi-(backend)
• docker
• elevenlabs
• extension
• face
• fastapi
• github
• html
• httpx
• hugging
• inference
• javascript
• orjson
• pydantic
• python
• python-dotenv-frontend-tools:-shadow-dom
• requests
• service
• shadow-dom
• sshleifer/distilbart-cnn-12-6)-speech-synthesis:-elevenlabs-text-to-speech-api-libraries:-pydantic
• uvicorn