Inspiration
Students often use different tools for different parts of learning: one for asking questions, another for reading notes, another for watching lectures, and another for practicing quizzes. We wanted to bring these experiences together into one personalized learning companion.
The idea behind LearnSphere AI was to create an AI tutor that could understand different types of educational content and adapt the learning experience to the student.
What it does
LearnSphere AI is a personalized, multimodal AI learning companion.
Students can:
- Ask questions and receive AI-powered explanations.
- Choose a subject and learning level.
- Upload TXT, PDF, and PPTX study materials.
- Ask questions directly about uploaded material.
- Receive source-grounded answers with references such as slide or page locations.
- Upload lecture videos for Gemini multimodal analysis.
- Analyze educational images, diagrams, questions, and code.
- Generate interactive five-question quizzes.
- Take adaptive quizzes whose difficulty changes based on previous performance.
- Review their Learning History.
The adaptive learning system uses the learner's previous quiz score to determine the next level of difficulty. Lower scores lead toward core concepts and easier questions, while stronger performance leads toward more challenging questions.
How we built it
The application uses a Next.js, React, and TypeScript frontend with a Python FastAPI backend.
Google Gemini provides the AI capabilities, including:
- Educational question answering
- Quiz generation
- Image understanding
- Multimodal lecture video analysis
- Source-grounded responses from uploaded learning materials
For documents, LearnSphere AI extracts educational content from uploaded files. PDF content is organized by page and PPTX content by slide, allowing the application to provide more useful source references when answering questions.
For lecture videos, Gemini's multimodal capabilities are used to analyze the video and produce structured educational content with approximate timestamps.
Challenges
One of the main challenges was combining different types of learning content into a consistent learning experience.
Documents, slides, images, and videos require different processing approaches. We also wanted answers based on uploaded material to remain connected to the original source, so students could see where supporting information came from.
Another challenge was implementing adaptive quizzes so that the next assessment responds to the learner's previous performance instead of always using the same difficulty.
What we learned
Building LearnSphere AI helped us understand how multimodal AI can be applied to real educational workflows.
We learned how to:
- Build an AI-powered full-stack application.
- Connect a Next.js frontend with a FastAPI backend.
- Work with Google's Gemini API.
- Process different educational file formats.
- Build source-grounded AI interactions.
- Use multimodal AI for image and video understanding.
- Implement adaptive assessment logic.
- Design a responsive learning interface.
- Test AI features through real learning scenarios.
Why LearnSphere AI
LearnSphere AI is designed around one simple idea:
Learning should adapt to the learner, not the other way around.
By combining multimodal study materials, source-grounded AI tutoring, and adaptive assessments, LearnSphere AI aims to make studying more interactive, personalized, and easier to understand.
Log in or sign up for Devpost to join the conversation.