Panorama Solara + Class Companion
AI platform for teachers to create assignments and students to get instant rubric-aligned feedback.
I work on agentic AI products at Panorama Education with a focus on site reliability and accessibility.
At Panorama Education I’ve recently been working on accessibility audits and setting up observability for my team’s new AI products. Recently I’ve also been working on adding granular o11y into steps of our durable workflows and agent skills to help with the tedious parts of a11y audits.
Outside of work I have been vibe coding to become more fluent with agentic development tools. I looked for any excuse to start a project and try out different code agents to both plan and implement ideas. I spent a large amount of time working through the design, architecture and decisions on the projects, but intentionally did not read most of the generated code. This has helped me to “read between the lines” when working with code agents and to guide code assistants through implementation, validation and debugging.
2026
WCAG 2.2 AA auditing that drives real screen readers (VoiceOver, Orca), axe-core, and an LLM agent through one workflow. Includes an eval harness that scores precision and recall against W3C ACT Rules.
AI-assisted analysis tool for Counter-Strike 2 demos. Upload a match, ask questions in natural language, watch annotated round replays, and inspect position and utility heatmaps. Match data and processing is handled in the browser, so match data is kept private aside from LLM calls.
Multi-tenant React and Supabase application to manage alliances in the Kingshot mobile game. ~25 users supporting 1,400 players. Auth, row-level security, edge functions, and OCR-powered score extraction. Supports mobile and desktop.
2025
Chrome extension that extracts Vue component context for AI coding agents. Click a component in your running app and get props, state, stores, routes, and a DOM locator formatted for Claude Code or Cursor. Inspired by react-grab.
An LLM evaluation platform that asks whether frontier models know what they don't know. Built for the Vercel AI Gateway Game Hackathon (Model Eval category).