Resume
Next.jsNext.js
TypeScriptTypescript
Tailwind CSSTailwindCSS
MySQLMySQL

Qaiz

Learn anything with AI-powered quizzes
Project Gallery
ProjectDetails.qaiz.screenshots.landing.title

Landing Page

The landing page showcasing the platform's features and benefits.

ProjectDetails.qaiz.screenshots.dashboard.title
ProjectDetails.qaiz.screenshots.dashboard.title
ProjectDetails.qaiz.screenshots.dashboard.title

Dashboard

The dashboard where users can create and manage quizzes.

ProjectDetails.qaiz.screenshots.quiz.title
ProjectDetails.qaiz.screenshots.quiz.title
ProjectDetails.qaiz.screenshots.quiz.title
ProjectDetails.qaiz.screenshots.quiz.title

Quiz Interface

The interactive quiz interface where users take quizzes and review their answers.

Overview

An innovative AI-powered quiz platform that generates personalized quizzes on any topic, helping users learn through interactive questions and detailed explanations.

Key Features
  • Generate quizzes on any topic using Gemini AI
  • Google OAuth for authentication
  • Topic validation using AI to filter gibberish/inappropriate content
  • Multiple-choice questions with a single correct answer and three plausible distractors
  • Post-quiz review with brief explanations for each question
  • Word cloud visualization of most frequently generated topics
  • History of past quizzes with statistics
Challenges & Solutions

Ensuring Valid JSON from Gemini Outputs

Gemini often returns responses with syntax issues, especially when generating JSON with nested structures or dynamic placeholders.

Solution: Implemented a robust generation function with retry logic, format constraints, and post-processing cleanup. Added strict validation rules for required fields, placeholder replacement, and sanitation of quotes and trailing commas.

Generating Plausible Distractors for MCQs

AI tends to repeat similar incorrect answers or mix up syntax when generating plausible MCQ options.

Solution: Fine-tuned prompt instructions with constraints: all options must be unique, under 15 words, and use single quotes. Embedded these directly into the generation logic to enforce consistency and relevance.

Preventing Garbage Input Topics

Users could input invalid, offensive, or nonsensical topics which led to low-quality quizzes or model confusion.

Solution: Added an AI-powered topic validation layer before quiz generation. It filters inappropriate or low-signal input and guides users to refine their topic.

GitHub