Tanya George

Curriculum Auditor finds the prerequisites a university course catalog gets wrong. As a grad student, I've had to read prerequisite lines and work out what I actually needed. Catalogs write every requirement as a sentence ("One of CS 173, MATH 213; CS 225"), so nobody sees the full structure, and mistakes can sit there for years: courses that require themselves, prerequisite loops, and requirements on courses that were renumbered or removed. One broken line can block a student at registration, and checking a whole catalog by hand means reading thousands of entries. I built a pipeline where Muse Spark reads each prerequisite and turns it into structured AND/OR logic, and code checks everything that has to be exact. Every course Muse Spark cites must appear in the original text; if one doesn't, it gets the specific error back for one retry, and anything still wrong goes to human review. The code then builds the full prerequisite graph and audits it, and every finding quotes the catalog text it comes from. To make sure it actually works, I hand-labeled 150 courses from 13 universities and held them out until the end. Muse Spark parses 94.7% of them exactly right, compared with 80.7% for a rule-based parser (94.7% vs 66.7% on the hard cases), at about $0.0025 per course. I also planted 400 known defects in real catalogs, and the audit found 99–100% of them with zero false alarms. The live app covers 21,739 courses across 14 universities, including UIUC's full Spring 2026 file parsed by Muse Spark. I built it solo during the hackathon. Live app: https://curriculum-auditor-seven.vercel.app