Skip to content
EduTrack

EduTrack

Personal AI-Based Study Guide

Created on 29th June 2025

EduTrack

EduTrack

Personal AI-Based Study Guide

The problem EduTrack solves

🧠 THE PROBLEM
Students often:

Struggle to make notes from long videos or PDFs

Waste time searching for quizzes or practice questions

Don’t get personalized feedback on what they’re weak at

Can’t visualize concepts clearly — especially in subjects like physics or CS

🚀 WHAT WE’RE BUILDING
We’re creating a smart learning platform where students can:

📥 Submit Notes or YouTube Links
Upload handwritten notes (PDFs) or paste in YouTube video links

Our backend parses them and automatically generates:

🧠 AI-Generated Learning Content
📑 Clean, structured notes

❓ Quizzes with multiple-choice questions and solutions

📊 Grading + weak area detection

🎯 Suggested resources for improvement (videos, articles, etc.)

🎮 Simulations (Unity)
To make topics truly understandable, we’re adding interactive Unity simulations

Concepts like projectile motion, pendulum, or even sorting algorithms

All loaded dynamically through Flask, no bloat, no separate builds

Challenges we ran into

🧩 Challenges We Ran Into
🔗 Unity WebGL + Flask Integration
We faced some issues getting Unity WebGL builds to run properly inside our Flask app.

Challenge: Making sure relative paths and static file serving worked for Unity’s Build/ and TemplateData/ folders.

Fix: We explored two approaches — treating it as a static asset vs rendering inside a Jinja template. Eventually, we served the WebGL build via Flask’s static folder with query-based scene switching using C#.

💻 Scene Management in Unity (C#)
Since we had multiple simulations (e.g., projectile, pendulum), loading different scenes externally was tricky.

Challenge: Unity doesn't support scene loading from outside very intuitively in WebGL.

Fix: We wrote a custom loader script in C# to read scene names from the URL query string and load them at runtime.

🧠 AI Quiz Generation
Challenge: Generating accurate and relevant quizzes from raw notes or transcribed video content is hard.

Fix: We created a simple pipeline using language models, but ran into quality and formatting issues (e.g., broken options, mismatched answers).

Discussion

Builders also viewed

See more projects on Devfolio