1. Multimodal Interface Auditing
Combines visual layout analysis with DOM trees to detect low-contrast elements, non-standard touch targets, and dynamic focus state issues that static linters miss.
Seaside Labs
Autonomous AI for software and web accessibility
We are building an intelligent, open-core accessibility engine that helps open-source maintainers and developers automatically detect, audit, and fix WCAG accessibility barriers directly in their codebases.
The Mission
Making software universally accessible to everyone
Over 95% of top websites and thousands of open-source projects fail fundamental WCAG 2.1 and 2.2 accessibility standards. For independent developers and open-source projects, manual audits and accessibility consulting are too cost-prohibitive.
Traditional accessibility tools rely on static regular expressions and basic DOM attribute checks. They catch obvious missing alt tags, but completely miss semantic navigation flow, keyboard traps, color contrast against dynamic backgrounds, and screen-reader comprehension.
Seaside Labs is solving this with an autonomous AI developer agent that understands visual layout, DOM hierarchy, and source code semantics, generating actual pull requests and automated fixes.
How our AI Accessibility Engine works
2. Semantic AST & Code Analysis
Parses component trees across modern frontend frameworks (React, Vue, Svelte, plain HTML/CSS) to trace ARIA relationships, heading structures, and keyboard navigability.
3. Autonomous Code Remediation
Rather than simply reporting errors, the engine generates ready-to-merge GitHub pull requests with clean, standards-compliant fixes, complete with test coverage.
4. Assistive Emulation
Simulates real-world screen reader output (NVDA, JAWS, VoiceOver) and keyboard-only navigation paths to verify compliance before deployment.
5. Open-Source CI/CD Integration
Plugs natively into GitHub Actions and continuous integration pipelines, preventing new accessibility regressions from being merged.
6. Privacy & Local First
Designed with developer privacy at the core. Compatible with open-weights models and private cloud instances so code never leaks to public training sets.
Technology & Cloud Architecture
Built for speed, accuracy, and scale
Our research and engineering stack blends large language models, computer vision, and high-performance serverless runners.
AI & Inference Layer
Powered by Azure OpenAI Service, multimodal vision LLMs, and fine-tuned open-weights models (Phi-4 / Llama 3) for deterministic WCAG rule verification.
Headless Auditing Runners
Distributed container runners running Playwright and headless Chromium, capturing DOM snapshots, accessibility trees, and visual renders in milliseconds.
Developer Experience
CLI tools, GitHub App integrations, and automated PR commentary that explains the accessibility failure in plain English alongside the patch.
Standards & Benchmarks
Benchmarked against WCAG 2.1/2.2 AA & AAA criteria, Section 508, and the European Accessibility Act (EAA).
Alpha Testing & Open Source Access
We are actively testing with select open-source maintainers and developer teams.
Free for eligible open-source repositories and accessibility contributors.
- Automated WCAG 2.2 audit report for your repository
- Initial automated PR with recommended accessibility patches
- Direct collaboration with our engineering team
- Help shape the open-core roadmap and developer tooling
Led by engineers passionate about open tech
Seaside Labs is the R&D software division of Seaside IT, founded by Dylan Roberts — a UK-based engineer and developer exploring the frontier of AI agents, open models, and digital accessibility tooling.
View Dylan's developer projects and research portfolio at droberts.xyz.