Inspiration
Learning today is mostly static — everyone gets the same explanation, the same quiz, and the same pace, regardless of how they actually learn.
We were inspired by two powerful observations:
- Great teachers adapt in real time when a student struggles or excels.
- Great games keep players engaged by adjusting difficulty, providing feedback, and rewarding progress.
We asked ourselves:
What if learning felt like a game, and the teacher adapted instantly like a human mentor?
That idea became LevelUpED — Learn like a game. Master like a pro.
What it does
LevelUpED is a real-time AI-powered learning platform that turns education into an interactive game.
- Each subject is a world
- Each concept is a quest
- Each assessment is a boss fight
- The AI acts as a live coach, not just a chatbot
The AI observes how learners study — their answers, speed, focus, and mistakes — and adapts the teaching strategy instantly.
Learners earn XP, unlock levels, and visualize mastery while genuinely improving their skills.
How we built it
We built LevelUpED as a Web + Mobile Progressive Web App (PWA) with a modular and scalable architecture:
Frontend
- React for UI and state management
- Phaser.js (HTML5 game engine) for gamified learning
- PWA support for mobile and offline readiness
Backend
- Node.js + Express
- PostgreSQL for structured learning, session, and analytics data
AI Layer
- Gemini API for:
- Concept explanations
- Quiz generation
- Storytelling
- Motivation and coaching
- Role-based prompts (Teacher, Quiz Master, Game Narrator, Coach)
The system continuously feeds user behavior back into the AI to create a real-time adaptive loop.
Challenges we ran into
1. Avoiding a “chatbot-in-a-game” trap
It was challenging to ensure the AI felt like a coach, not just a Q&A bot.
Solution:
We split AI behavior into multiple roles and tied AI decisions directly to gameplay signals.
2. Balancing scope with execution
Gamification, AI adaptation, analytics, and PWA — all within limited time.
Solution:
We clearly separated:
- MVP features (implemented)
- Advanced features (roadmap)
3. Making AI decisions trustworthy
Adaptive AI can feel arbitrary if users don’t understand why it changes behavior.
Solution:
We logged AI interventions and connected them to observable user actions like hesitation or repeated mistakes.
Accomplishments that we're proud of
- Built a real-time adaptive learning system, not a static tutor
- Successfully merged education + gaming + AI into one cohesive experience
- Designed a multi-world learning structure that scales across subjects
- Created measurable learning metrics (focus, accuracy, improvement)
- Delivered a product that feels engaging, intelligent, and purposeful
What we learned
- Learning is behavioral, not binary — time and hesitation matter as much as correctness
- AI is most impactful when it adapts, not just responds
- Gamification works best when it reinforces mastery, not distraction
- Clear system design and prompt architecture are as important as model capability
- Showing both execution and vision builds credibility
What's next for LevelUpED — Learn like a game. Master like a pro
This is just the beginning. Our roadmap includes:
- Voice-based learning and explanations
- AI coach personality selection
- Multiplayer and co-op learning modes
- Emotion-aware and focus-aware coaching
- Institution and classroom dashboards
- Skill-to-career mapping and certifications
Our vision is to make LevelUpED the learning operating system for the next generation — where every learner has a personal AI coach and learning feels as engaging as the best games.
🚀
Built With
- bun
- chrome
- express.js
- gemini
- html5
- node.js
- phaser.js
- postgre
- prisma
- react
- typescript
Log in or sign up for Devpost to join the conversation.