PhotoShelter: Accessibility Leadership & Product Prototyping

Accessibility
PhotoShelter's customer base includes universities and professional sports organizations, and several key contracts worth a combined $9.2M ARR were exposed if the product could not demonstrate ADA and WCAG compliance. I led the WCAG 2.1 audit and front-end remediation effort for PhotoShelter's core product to close that gap.
The initial audit surfaced two classes of issues: fast fixes such as missing ARIA attributes and unlabeled controls, and deeper structural problems such as mouse-only navigation across core menus.
Keyboard navigation for every menu
Keyboard navigation for the main menu
Before this work, every menu in the product was mouse-only. Keyboard access is essential for users with limited mobility and speeds things up for power users who prefer shortcuts.
The hardest part was keyboard support across Shadow DOM-based components. Because PhotoShelter relied on Polymer and Web Components, focus and keyboard events did not move cleanly across shadow boundaries. I built a custom layer to bridge those boundaries so interactive elements remained reachable and usable without switching interaction modes.
Mouse to keyboard control, no mode switch needed
Team adoption
I mentored four developers on accessibility patterns and introduced review checklists so accessibility was built into new work rather than added later as remediation.
R&D Prototyping
In parallel, I worked with the Chief Architect on exploratory prototypes outside the main production timeline. These projects helped test ideas and catch integration problems before they reached production.
Workflow automation interface
This prototype informed the Smart Galleries feature. Building it outside the production path surfaced integration issues earlier and reduced development time by roughly 30%.
Client-side image blur comparison
Built with OpenCV.js to test whether image analysis could run fully in the browser without backend resources. The prototype showed that lightweight computer-vision workflows were viable in static-file environments, even though it did not ship as a standalone feature.