ai-quiz-generator

"An AI-powered smart-tutor that transforms any document into an interactive, highly-customizable quiz with real-time feedback and dynamic learning analysis."

Comment

💡 Inspiration

Traditional learning often relies on passive reading with no real-time guidance. We wanted to build a "smart tutor" that goes beyond testing your knowledge and actively teaches you as you practice, gently correcting mistakes exactly the way a human teacher would.

💻 What it does

Our app transforms any text into a customized, interactive educational experience. Users configure their difficulty, question count, and topic focus.

As you take the quiz, it acts as an intelligent tutor:

  • The Real-Time Teacher: The moment you select an answer, an animated box dynamically explains why the answer is right or wrong, rectifying misconceptions immediately.
  • Intelligent Post-Quiz Analysis: After the quiz, it correlates your answers against hidden core concepts. It outputs a color-coded Learning Analysis, providing "Tutor Cards" that serve as a bespoke study guide highlighting your strengths and weaknesses.

⚙️ How we built it

The frontend was built entirely in Vanilla HTML, semantic CSS, and JavaScript to remain lightweight, utilizing a premium "glassmorphic" aesthetic with animated mesh gradients.

The backend intelligence is driven by the Google Gemini-Flash API. Using advanced prompt engineering, we force the AI to return perfectly structured JSON data—including custom parameters for concepts and explanations—which our JavaScript parses dynamically to build the UI on the fly.

🚧 Challenges we ran into

Consistently extracting the exact JSON structure from the AI was tough. Stray Markdown characters often broke our frontend loops. We overcame this by tightening our prompt constraints and implementing rigorous Javascript regular expression sanitizers before parsing the data.

🏆 Accomplishments that we're proud of

We seamlessly interwove the "testing" and "teaching" phases. Watching the AI automatically infer the underlying concept of an incorrect answer and immediately formulate an encouraging "Tutor Card" on the fly feels incredibly futuristic.

🧠 What we learned

We vastly improved our skills in asynchronous JavaScript (async/await) and deepened our understanding of Prompt Engineering—specifically how rigid instructions must be when commanding an AI to output code-ready data instead of conversational text.

🚀 What's next

We plan to implement native PDF and .docx parsing and integrate local browser storage to track student progress over time, generating evolving, multi-week study plans automatically.

Share this project:

Updates