Following the success of the Neuro Nostalgia Hackathon that closed out in 2024, Hackathon Raptors has completed its second major event of the year – the AI-Powered DEI Web Accessibility Hackathon. This 72-hour virtual competition, held from February 28 to March 3, 2025, brought together developers worldwide to tackle one of technology’s most pressing challenges: making the web accessible to all users through AI-enhanced solutions.
This event builds upon Hackathon Raptors’ growing reputation for creating signature technical competitions, continuing the momentum from their previous IAHD DEI Web Accessibility Hackathon held in September 2023. That 48-hour event focused on e-commerce and retail accessibility challenges, laying the groundwork for this expanded AI-powered iteration. While the 2023 hackathon emphasized visual accessibility, cognitive accessibility, text resizing, and keyboard accessibility, the 2025 event pushed participants further by requiring sophisticated AI implementations to address these same challenges at scale.
The Technical Challenge
The hackathon addressed a critical gap in digital infrastructure: according to WebAIM’s 2023 analysis, 96.8% of the top million website homepages contain detectable WCAG failures. As the global AI in accessibility market projects to reach $7.3 billion this year (MarketsandMarkets, 2023), participants were tasked with creating functional, AI-powered accessibility tools that could be deployed in real-world applications.
Unlike traditional hackathons focusing on proof-of-concept demonstrations, this event required participants to build solutions addressing specific technical requirements:
- Ethical AI implementation principles
- Resource optimization for performance
- WCAG 2.2 compliance through AI assistance
- Cross-platform compatibility with scalable architecture
- Implementation of current AI technologies (LLMs, computer vision, speech processing)
Technical Deep Dive: Winning Projects
VoxSurf (1st Place)
Team VoxSurf captured first place with a comprehensive desktop application built specifically for visually impaired users, featuring:
- Neural speech-to-intent processing that converts natural language commands into precise browsing actions
- Custom-trained computer vision model that identifies UI elements even when not properly tagged
- Context-aware RAG implementation that maintains user navigation history to improve accuracy
- Low-latency feedback system providing audio descriptions of visual content
The technical architecture leverages a client-side application that processes voice commands locally before sending complex requests to a cloud backend for heavier AI processing, maintaining responsiveness while providing sophisticated assistance.
Accessibility Helper (2nd Place)
Team Lakshan Krithik’s browser extension demonstrated innovative input methodologies with:
- Real-time facial feature tracking using optimized computer vision algorithms that function even in low-light conditions
- Machine learning model trained to distinguish between intentional and unintentional head/nose movements
- Predictive cursor positioning that anticipates user intent based on page context
- Hardware-accelerated processing for maintaining 60fps performance despite complex AI operations
The solution enables users with limited mobility to navigate web interfaces with minimal physical input, potentially transforming web accessibility for users with severe motor impairments.
Browsingforall (3rd Place)
This solution implemented:
- Voice recognition engine optimized for command detection with high accuracy even in noisy environments
- Computer vision system that analyzes rendered web pages to identify interactive elements
- Command execution pipeline that translates voice inputs to DOM interactions
- Accessibility notification system that proactively identifies potential navigation barriers
ComplyScan (Category Winner: Technical Architecture)
Team CtrlShiftHack’s compliance automation tool leveraged sophisticated technical approaches:
- Static and dynamic page analysis algorithms that evaluate keyboard navigation paths
- Contrast analysis engine using computer vision to verify color combinations against WCAG requirements
- Machine learning classifier to identify and flag non-compliant page elements
- Integration API allows the tool to work within existing development environments
Expert Technical Assessment
Igor Kiselev, Principal Director at Accenture and member of multiple professional organizations, including the IEEE, CFA, and Forbes Technology Council, evaluated projects based on technical architecture and scalability. With his extensive background in business applications and digital twin models, he provided critical insights on real-world deployment potential.
“What impressed me most about the winning solutions was their approach to system architecture,” Kiselev noted. “VoxSurf, in particular, demonstrated exceptional technical foresight by implementing a hybrid processing model that balances local and cloud computing resources. This approach ensures both privacy and performance—critical factors for accessibility tools that must function reliably across varying internet conditions. The team’s implementation of dynamic response handling based on network latency showed enterprise-grade thinking.”
Chingiz Mizambekov, Lead Functional Analyst and Proxy Product Owner at Qframe NV, brought his healthcare digitalization expertise to the judging panel. With over a decade of experience in software development and project management, including work on the Factura invoicing system for the University Hospital of Antwerp, Mizambekov evaluated projects for their practical implementation in regulated environments.
“The ComplyScan solution stood out for its technical thoroughness,” Mizambekov observed. “The team built a verification system that doesn’t just identify accessibility issues but provides contextual remediation suggestions with code samples—similar to approaches we’ve implemented in healthcare systems where compliance is non-negotiable. Their approach to programmatically verifying ARIA roles and semantic HTML structure demonstrates a deep technical understanding of both accessibility standards and modern web architecture.”
Technical Implementation Challenges
Participants navigated several significant technical hurdles throughout the competition:
- AI model optimization: Balancing model size and accuracy for browser-based deployment without compromising performance
- Cross-device compatibility: Ensuring consistent behaviour across operating systems, browsers, and assistive technologies
- Real-time processing: Minimizing latency for accessibility tools that require immediate feedback
- Ethical considerations: Addressing potential AI biases while maintaining user privacy and data security
Future Technical Directions
The hackathon highlighted several promising technical approaches for improving web accessibility:
- Hybrid AI architectures that combine edge and cloud processing for optimal performance
- Multimodal interfaces that adapt to user capabilities rather than requiring users to adapt to interfaces
- Automated compliance verification integrated directly into development workflows
- Contextual assistance that understands user intent beyond literal commands
As organizations face increasing legal and ethical pressure to make digital experiences accessible to all users, the technical solutions demonstrated at this hackathon represent significant advancements in addressing the complex challenges of web accessibility.
With the remarkable success of the AI-powered DEI Web Accessibility Hackathon, Hackathon Raptors has solidified this event as a cornerstone in the technical accessibility community.
Given the extraordinary innovations demonstrated this year, plans are underway to expand the event in 2026 and 2027, potentially incorporating multi-stage development cycles and enterprise implementation tracks. As digital accessibility continues evolving at the intersection of AI and inclusive design, this hackathon is positioned to become the definitive annual benchmark for technical progress in making the web accessible to all users.
