Skip to Main Content

AccessiWebEngine

Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

AccessiWebEngine — Demo video

THE PROBLEM The generative AI boom has accelerated web development velocity to unprecedented levels, creating massive "accessibility debt." Over 96% of web homepages fail basic WCAG 2.2 accessibility compliance, locking out millions of users with visual, motor, auditory, or cognitive disabilities. Existing developer tools fall into two broken extremes: static linters that generate noisy warnings without fixing code, or manual audits that are too slow and expensive for modern CI/CD pipelines. THE SOLUTION AccessiWebEngine is an autonomous, closed-loop web accessibility auditing and code remediation platform powered by the Meta Model API, Muse Code, SAM 3.1, Muse Spark Open, Muse Wonder, and Muse Image. It bridges visual UI perception and source-code AST refactoring across three unified experiences: 1. Live Web URL Accessibility Analysis (Public): Users input any live website URL. The engine extracts the DOM, uses SAM 3.1 to visually segment interactive UI components (verifying touch target sizes, layout overlaps, and contrast boundaries), and uses Muse Code to generate a comprehensive WCAG 2.2 AA/AAA compliance report. It also uses Muse Wonder to synthesize an interactive "Screen Reader Empathy Audio Preview" comparing the broken vs. remediated experience. Users can copy the formatted report or dispatch the automated GitHub Bot to open a detailed GitHub Issue in 1 click. 2. Public Repository Source Code Analysis (Public): Users input a public GitHub repository URL. Muse Code performs deep static AST analysis across frontend files (.tsx, .jsx, .html, .vue, .svelte, .css), pinpointing exact file paths, line numbers, WCAG 2.2 criteria violated, and side-by-side code diff recommendations—with 1-click automated GitHub Issue creation. 3. Autonomous Remediation & Pull Request Bot (Authenticated Preview Mode): Multi-step autonomous agent (Scan -> Plan -> Patch AST -> Verify -> Document -> Open PR) uses Muse Code to refactor offending files across the repository, generates a full `/docs/ACCESSIBILITY_REMEDIATION.md` audit log, and submits a ready-to-merge GitHub Pull Request automatically. ENTERPRISE CLOUD ARCHITECTURE Built as a TypeScript monorepo deployed on Azure Static Web Apps, Azure Functions, Azure Cosmos DB, and Azure Key Vault, complete with SSRF and prompt-injection defenses, modular subscription-scoped Azure Bicep IaC, automated ephemeral PR environments via GitHub Actions, and Playwright + Axe-core E2E tests verifying zero accessibility violations on our own platform.

Team