AIMRoyale

Inspiration

School systems haven't evolved in decades, and private tutors are way too expensive for the average student. I wanted to use my summer vacation to build something meaningful for my portfolio that could democratize high-level learning. The goal was to bridge the gap between rigorous academic preparation (like math competitions and exams) and engaging internet/space culture. Why can't a physics problem be about Stephen Curry launching a three-pointer from a NASA lunar base? AIMRoyale was born from the idea that education shouldn't be a boring lecture—it should be a gamified arena.

What it does

AIMRoyale is a free, gamified space-math arena that aligns directly with actual school curricula and science chapters (such as Universal Gravitation or Quadratic Functions). Powered by Alibaba Cloud's Qwen-Max model, it dynamically generates 100% custom problems tailored to the user's grade, class, and skill level.

Instead of generic quizzes, students navigate through high-stakes space scenarios. The platform features an intelligent, privacy-first memory agent that tracks specific academic weaknesses cross-session. It turns studying into an interactive video game where players actively learn from their mistakes.

How we built it

I spent 5 hours a day orchestrating this system from scratch:

• Frontend: Built with standard JavaScript, HTML, and CSS, fully optimized and wrapped as an installable PWA (Progressive Web App) for mobile accessibility.
• Backend: Engineered a secure Node.js Express proxy deployed on Render to safely route requests to the /api/generate endpoint, keeping production-grade security intact.
• AI Core: Integrated Alibaba Cloud's API to leverage the reasoning power of the Qwen-Max LLM.
• State & Logic: Architected a zero-database context engine using a custom LocalMemoryAgent loop built on top of the Web Storage API (localStorage).
• Optimization: Wrote a custom Smart Decay Algorithm that automatically flushes old error logs out of the prompt context once a student hits a 3-answer consecutive winning streak, drastically reducing token consumption.

Challenges we ran into

• API Key Exposure: Initially, querying the LLM directly from the client exposed our credentials in the network tab. I had to learn how to build and deploy a backend proxy to completely obfuscate the Alibaba Cloud API Key.
• Database Fatigue: I experimented with integrations like Firebase but realized I wanted a zero-login, privacy-first, instantaneous onboarding experience. Overcoming this meant pushing localStorage to its absolute limits to handle multi-session structured learning telemetry.
• Context Bloat: Raw error logging quickly clogged the prompt context, making queries slower and expensive. Designing the logic for the Smart Decay Algorithm to seamlessly "forget" mastered concepts was a heavy math and prompt engineering hurdle.

Accomplishments that we're proud of

As a 16-year-old solo developer, I am incredibly proud of executing a full-stack, production-ready architecture in just a week of summer vacation. I managed to build an advanced, context-aware AI coach that adapts to user mistakes without needing a massive, expensive server-side database. Successfully packaging the engine as a functional PWA that runs perfectly on mobile is a massive milestone for my coding journey.

What we learned

I gained a ton of real-world, professional experience. I learned how to treat LLMs as dynamic state engines rather than simple text generators by feeding them structured local data. More importantly, I discovered how to leverage AI as a force multiplier, using it to accelerate my learning.

What's next for AIMRoyale

This is just the baseline infrastructure for a global educational platform:

• Rich Media: Automatically generating and attaching tailored meme images to every single problem created by the engine.
• Gamification: Implementing a fully open-source peer-to-peer leaderboard and trophy system.
• Deep Tracking: Next year, I plan to completely rewrite this project from scratch to implement advanced AI telemetry capable of analyzing long-term learning rhythms.
• Community First: Keeping the repository 100% open-source so anyone, anywhere, can deploy it to help kids who can't afford private education.

Built With

• alibaba-cloud-api
• css
• express.js
• html
• html5
• javascript
• localstorage-api
• node.js
• pwa
• qwen-max-model
• render