A 5,000+ course catalog with a multi-year QA backlog and only 40% audit coverage. Manual per-course audit takes ~5 hours.
Architected an automated evaluation pipeline using Codex, Claude, and Python — async workers, job queues, batching, retry logic. Grades against WCAG 2.1 AA and instructional-design rubrics; surfaces findings to reviewers.
Per-course audit time: 5 hours → 15 minutes (~95% reduction). Tested end-to-end on the 5,000+ course catalog architecture.
Building a course from existing curriculum materials takes weeks of faculty + ID time, even when the curriculum itself is already written.
Built an AI-assisted authoring plugin that consumes curriculum materials and generates Canvas-ready course shells — content, assessments, accessibility checks, ID-standards validation — in a single workflow.
Build time per course (with all curriculum materials supplied): weeks → 30–60 minutes. Per-course edit time: 6 hours → 10 minutes. Validated end-to-end.
Course review → fix → verify cycles consume thousands of labor-hours per year across EdPlus.
End-to-end pipeline: automated audit findings flow into remediation queues with verification checks on the way out. Closes the loop between detection and fix.
Per-course review-fix-verify cycle reduced ~70% in tested workflow. Automated audit findings flow into remediation queues with verification checks on the way out.