The problem HORCRUX solves

The Problem it solves

Traditional STEM education suffers from a massive engagement crisis. Students are often subjected to static, repetitive quizzes and one-size-fits-all curricula that fail to spark curiosity. The material is dry, question banks run out of fresh content, and students often feel lost regarding how their current studies connect to a future career.
Horcrux solves this by transforming passive study into a high-stakes, multiplayer Wizarding World adventure.

It addresses three core problems:

The "Static Content" Problem: unlike standard quiz platforms that rely on fixed databases, Horcrux uses Generative AI (Groq + Llama 3) to generate unique, context-aware questions instantly from any topic or uploaded PDF. This means the learning material is infinite and never repetitive.

The "Boredom" Problem: We replaced solitary revision with social competition. By using Socket.io for real-time 1v1 Duels and Squad Battles, we turn learning into an active, competitive sport where speed and accuracy determine the victor.

The "Directionless" Problem: Students often don't know what to do with their skills. Our AI Career Guidance acts as a multi-agent system that analyzes a student's grades, interests, and quiz performance to generate personalized career roadmaps, bridging the gap between today's lessons and tomorrow's jobs.

In short: We make education addictive by replacing boredom with magic.

Challenges we ran into

Challenges I ran into
Building a real-time multiplayer platform with generative AI integration as a beginner.

Synchronizing Game State via WebSockets: One of the hardest parts was ensuring that both players in a 1v1 duel saw the exact same question at the exact same time. Early on, we faced race conditions where one player would see the next question while the other was still on the previous one. We solved this by centralizing the game clock on the server (Node.js) and broadcasting "state snapshots" to clients rather than letting clients manage their own timers.

Taming the AI (Prompt Engineering): Initially, the Llama-3 model would sometimes return questions with broken JSON formatting or hallucinations, which crashed our frontend. We had to iterate extensively on our "System Prompts" to enforce strict JSON schemas and added a robust error-handling layer that cleans and validates the AI's output before sending it to the game engine.

Persisting User Progress Without a Database: To keep our architecture lightweight for the hackathon, we avoided setting up a heavy database like MongoDB. However, we still wanted users to track their monthly progress. We implemented a persistent local storage strategy that syncs seamlessly with the React state, allowing the "Monthly Progress" bar and battles "History" to update dynamically and survive page refreshes.

The "Merge Conflict" Nightmare: Collaborating on the same React components (like Dashboard.js and App.js) simultaneously led to some intense merge conflicts. We learned the hard way to communicate better before pushing changes!

These challenges ultimately pushed us to write cleaner, more modular code and gave us a deeper understanding of real-time architecture.