Project 02
SmartQuiz
AI-assisted quiz and TOS system
In development
A capstone system that helps educators write quizzes, generates the Table of Specifications (TOS), and routes exams through faculty review. A fine-tuned DistilBERT model sorts questions by Bloom's Taxonomy level.
- Role
- Full-stack developer
- Team
- Team of 2, sharing one system with a second capstone team
- Year
- 2026
- Stack
- React 19
- Vite
- Tailwind CSS v4
- Supabase
- FastAPI
- PyTorch
- DistilBERT
- Source
- The repository belongs to my capstone partner, so it isn't linked here.
Problem
Writing an exam means balancing questions against syllabus objectives and difficulty levels, then building a Table of Specifications (TOS) by hand and passing it through review. SmartQuiz automates the TOS and supports review in several tiers.
My role
Full-stack developer on a team of two. Our team and another team defended the same capstone title, so the school paired us: we build one shared system and write separate papers.
Inside that system I built our team’s features, and I also built every role’s side of the app: instructor, senior faculty, department head, and the student side for taking exams.
My partner built the audit trail, the question bank, student profiles, subject requests, and the item analysis that uses Gemini to suggest better questions and answers.
What I built
- The instructor, senior faculty, department head, and student exam-taking sides of the system.
- A fine-tuned DistilBERT model, served through FastAPI, that classifies quiz questions into Bloom’s Taxonomy cognitive levels.
- The TOS generation and curriculum alignment module, which balances exam blueprints against syllabus objectives and difficulty.
- Authentication, roles for Instructor, Reviewer, and Admin, PostgreSQL Row Level Security policies, and Cloudflare Turnstile bot protection.
Status
In development. There is no public demo yet.
Screens
Screens show development test data.
