Overview
An AI-assisted course-planning and mastery-based-practice platform for university instruction, advised by Prof. Diba Mirza. I own two pieces of it: a project-management ticket tracker that coordinates a 20+ person team across 5 concurrent Caliber projects, and a redesign of the LeetCode Autograder’s student submission flow and feedback interface.
Tracker launching in the coming weeks; the autograder redesign launches fall 2026 in UCSB’s CS8 and CS24 (250+ students/quarter). Contributed to platform work published at ACM ITiCSE ’26.
The problem
The Caliber team had grown to 20+ people across 5 concurrent projects with no shared way to see who owned what. Work was getting tracked informally, which made it hard to tell what was in progress, blocked, or done. Separately, the LeetCode Autograder gave students a flat pass/fail with no breakdown of what actually went wrong.
The decision
We compared the tools the team already used (GitHub Projects for Kanban boards, Notion for dashboards), then prototyped our own tracker in Figma and built it. After demoing v1 at standup, we added project tags so one tracker could serve every project.
The outcome
A ticket tracker supporting creation, assignment, and self-claiming, with status, ownership, deadlines, and a change history — launching in the coming weeks for the whole 20+ person team. On the autograder side, restructured the 6-step grading pipeline into clear pass/fail breakdowns that separate what passed, what didn’t, and what to fix, ahead of the fall 2026 launch.
What I'd do differently
If I were to rebuild this, I would implement different views, such as Calendar or Gallery to view all tickets, as well as different visualizations for basic progress tracking. We maintained a consistent TA and student feedback loop and implemented/revised features as requested.