RubricGuard AI — A grading copilot universities actually trust.

The problem

University grading is high-variance, slow, and hard to audit. Instructors and TAs grade the same work inconsistently. Students don't trust the rubric. Re-grade requests pile up at the end of every semester. The cost is hidden — in instructor time, in student trust, in the academic record itself.

What I built

• — Took the product from zero to working MVP, currently in active user testing.
• — Ran ~20 educator and TA discovery interviews to validate the problem and shape the feature set.
• — Authored the PRD, defined user personas, prioritized features, sized TAM, and built the pricing model.
• — Designed the RAG architecture and agentic workflow that powers the grading engine.
• — Defined the pilot strategy with CMU's Eberly Center for Teaching Excellence & Educational Innovation.

Product gallery

Screens from the live MVP — landing, grading, analytics, fairness.

Discovery storyboard

A 9-panel storyboard built during discovery to align educators on the problem and the copilot's role.

Recognition

2nd Runner-up (3rd place), ProdHacks Hackathon '26 — Tepper School of Business at Carnegie Mellon University. ~50–70 participating teams.