QurioPilot: Revolutionizing Education Through AI-Powered Interactivity

Inspiration

We were frustrated by how traditional education still treats curiosity like a side quest, not the starting point. Students are expected to follow rigid curriculums, with little space to ask questions or explore tangents that genuinely interest them. What if we flipped that? What if a learner could simply ask "Why do chairs look so different across history?" and get an entire personalized learning journey?

That's how QurioPilot was born — a platform where curiosity drives the learning, not just supplements it.

What it does

QurioPilot transforms any natural-language question into:

A set of focused subtopics to explore
Concise explanations of each subtopic
Interactive activities (like drag-and-drop games, match pairs, quizzes) that make the learning fun and sticky

It's like having a personal tutor, content creator, and educational game designer — all rolled into one AI-powered platform.

How we built it

Frontend: React + TypeScript, styled with Tailwind and animated using Framer Motion. React Router handles smooth stateful navigation between curiosity paths.

Backend: FastAPI handles all logic, AI queries, and JSON responses.

AI Model: We used Meta's open-source LLaMA-3.3-70B-Instruct-Turbo-Free model for generating subtopics, explanations, and activities. It's instruction-tuned, lightning-fast, and runs cost-effectively at scale.

Pipeline: Every user question is passed through a GPT-style reasoning engine that:

Identifies the subject, depth, and subtopics
Chooses an interactive format (e.g., drag_drop, fill_blanks)
Generates JSON-ready content for the frontend to render games

Challenges we ran into

Unpredictable JSON from the AI: AI responses sometimes broke the frontend. We implemented layered retry logic, JSON repair functions, and prompt engineering tailored for LLaMA to fix this.

Maintaining context across flows: Preserving the original question through subtopic selection and explanation phases was tricky. We used React Router state passing + backend memory objects to keep things in sync.

Responsive drag-and-drop on all devices: Making interactions smooth on mobile and desktop required careful event handling, Framer optimization, and real-time feedback systems.

Accomplishments that we're proud of

Built a fully functional prototype that creates interactive learning experiences from any student question
Achieved JSON consistency and performance using open-source models (no expensive APIs!)
Designed an elegant frontend that makes curiosity exploration feel intuitive and fun
Enabled real-time interaction generation for any subject: psychology, physics, history—you name it

What we learned

Curiosity is underrated: Designing around it creates deeper learning and higher motivation
AI + interactivity = magic: Engagement skyrockets when students actively play with ideas, not just read them
Open-source AI is viable at scale: LLaMA-3.3 delivered competitive quality while keeping costs and flexibility in check
UX matters more than content: Even great explanations fail without flow, feedback, and smooth animations