Education technology
A human-led AI workspace for rubric-based assessment that prepares evidence, provisional marks, and feedback while keeping every academic decision with the instructor.
The challenge
Rubric-graded, artifact-based work can consume evenings of careful evidence finding, criterion-by-criterion judgment, and feedback writing. Generic AI tools can save time, but often hide the reasoning or weaken instructor control.
The product
GradeSteward links submissions to an instructor-approved rubric, calibrates the approach on one sample, and drafts criterion-level evidence, marks, and feedback for review and finalization.
See how workflow discovery becomes a focused productThe principle
“AI should support academic judgment, not silently replace it. Every mark remains provisional until the instructor reviews and approves it.”
A reviewable starting point for every criterion
Clear evidence behind provisional decisions
Instructor authority preserved from setup to final grade
My role
01
Problem discovery from first-hand teaching experience
02
Assessment workflow and product design
03
Human-in-the-loop AI architecture
04
Educator pilot design and validation
