Project
Rubric-Guided Grading Assistant GPT
A grading companion that helped faculty work through a rubric and draft feedback without handing over the final judgment—or the student’s paper.
Overview
This was built as a faculty-facing grading companion rather than an automated grader.
The goal was to make grading faster and more consistent without handing evaluative authority to a model.
What I built
It is a guided rubric workflow that turns broad criteria into focused evaluation prompts, with scoring support that shows the reasoning behind a judgment rather than just an answer, and draft feedback that stays editable and under the instructor's control.
Design posture
I treated trust and control as first-order requirements rather than add-ons. The system supports faculty judgment instead of bypassing it, which kept the prototype grounded in how grading really works and kept it from overclaiming what the tool should do.
Current state
This remains a prototype and a design pattern for privacy-conscious, instructor-led AI assessment.
Keep exploring