Inspiration
We noticed that while AI research tools like NotebookLM are powerful, they often fail to make STEM content accessible to visually impaired users. Mathematical equations, complex diagrams, and scientific figures remain inaccessible barriers in education and research.
What it does
Accessify transforms inaccessible PDF documents into screen-reader friendly content using AI-powered accessibility enhancement. It combines NotebookLM-style document interaction with automatic alt text generation, making complex mathematical content accessible while providing chat-based research capabilities.
How we built it
Frontend: React + TypeScript with Tailwind CSS and shadcn/ui Backend: Supabase Edge Functions (Deno) with PostgreSQL and vector search AI Integration: Mathpix API for PDF→HTML conversion, Google Gemini for accessibility enhancement Workflow Automation: N8N for document processing and AI chat functionality Accessibility: Semantic HTML and ARIA labels
Challenges we ran into
Async Processing: Ensuring the accessibility enhancement pipeline completed before returning success responses Gemini API Integration: Switching from REST API to proper SDK implementation for reliable AI text generation HTML Manipulation: Properly placing accessibility descriptions next to their corresponding images without bunching them up Timeout Issues: Managing long-running AI operations without premature function termination n8n Integration Issues
Accomplishments that we're proud of
Real-time Accessibility Scoring: Live metrics showing accessibility improvements during processing AI-Powered Alt Text: Gemini-generated WCAG-compliant descriptions for mathematical content Complete Pipeline: Seamless PDF→Mathpix→Gemini→Accessible HTML workflow Screen Reader Optimization: Mathematical equations rendered as readable text for assistive technologies
What we learned
AI API Reliability: The importance of using official SDKs over raw REST calls for production applications Accessibility Standards: Deep dive into WCAG 2.1 AA compliance and semantic HTML requirements Async Programming: Proper handling of long-running operations in serverless environments User Experience: The critical importance of clear progress indicators for AI processing workflows
What's next for Accessify
Multi-language Support: Extend accessibility features to documents in different languages Advanced Math Rendering: Integration with MathJax for better mathematical equation accessibility Collaborative Features: Real-time collaborative research sessions with accessibility features Enterprise Integration: API for organizations to batch-process large document collections Mobile Accessibility: Native mobile app with full accessibility compliance
Built With
- react
- supabase
- typescript
- vite

Log in or sign up for Devpost to join the conversation.