Orbita
Colosseum
"Sky is the limit"
Submitted
May 22, 2026, 8:05 AM
Last Updated
May 22, 2026, 8:05 AM
Project Links
Problem Statement
COLOSSEUM solves the problem of ineffective last-minute exam preparation. Most students face syllabus overload, scattered resources, lack of guidance, low motivation, and severe time pressure before exams. Existing study apps mainly focus on note-taking or passive learning, but they don’t help students decide what to study first, what to skip, or how to maximize marks in limited time. COLOSSEUM acts as an AI-powered “exam war room” that analyzes study material and generates a personalized high-impact strategy, including priority topics, predicted questions, cheat sheets, schedules, mock tests, and curated learning resources. It also uses gamification and an AI mentor system to keep students motivated under stress. This matters because millions of students struggle with exam anxiety and inefficient preparation, especially in short timeframes. COLOSSEUM transforms panic into a structured action plan.
Solution & AI Usage
We built COLOSSEUM using a modern full-stack architecture focused on speed, AI integration, and immersive UX. The frontend was designed with HTML, CSS, and JavaScript for an engaging “exam war room” experience. The backend handles AI workflows, scheduling logic, progress tracking, and personalized study generation. For AI, we used GroqAPI and npm multer language models to analyze uploaded PDFs, previous year papers, and syllabus data. The models were chosen because they are strong at summarization, topic prioritization, question prediction, adaptive planning, and conversational mentoring. We also integrated YouTube-based resource recommendations to provide fast topic-specific learning support. The system combines AI reasoning, gamification, and productivity design to convert unstructured exam stress into a personalized high-impact preparation strategy.
Full Description
COLOSSEUM — AI Exam Intelligence Engine
Upload your university's previous year papers, tell us how many hours you have left, and Colosseum predicts what will actually come in your exam — then builds a survival strategy in under 2 minutes.
The Problem
Students don't fail because content is unavailable. They fail because there's too much of it and zero prioritization. ChatGPT explains concepts. YouTube explains concepts. Nobody tells you what's actually coming in your university's exam tomorrow.
VTU sets numerical-heavy papers. MAKAUT repeats questions. Anna University wants definitions. JNTU loves derivations. Generic AI is blind to this. Colosseum isn't.
What It Does
Upload up to 5 PYQs → enter your subject, university, hours left → get a full battle plan in 2 minutes:
Probability Heatmap — every topic ranked by likelihood with reasoning ("Deadlock: 92% — appeared 4 of last 5 years") Predicted Question Paper — AI-generated paper matching your university's exact format Hour-by-Hour Schedule — exactly what to study and when Emergency Mode — if you have ≤3 hours, only survival topics, maximum ROI Interactive Lessons — tap any hour slot for a full AI lesson with memory tricks, diagrams, and YouTube videos Self Exam + AI Grading — attempt the predicted paper, get scored and feedback instantly War Room — focused cheat-sheet study with a live timer Senior Bhaiya — AI chatbot that talks like your brilliant final-year senior Points & Rewards
Study minutes, completed topics, and exam scores earn points. Points redeem for real canteen credit (₹10/₹25/₹75) via QR code. A live leaderboard ranks students by points.
Tech Stack
Node.js + Express backend, Vanilla JS frontend, SQLite database, Groq API (Llama 3.3 70B), YouTube Data API, pdf-parse for PYQ extraction. Deployed on Render.
Why It's Different
Most hackathon AI projects are chatbots with a system prompt. Colosseum is a full-stack intelligence engine with university-specific pattern detection, real authentication, a gamification layer, and a working deployment. The problem is real, the demo is visceral, and every engineering student in the room instantly gets it.
Tech Stack
Screenshots
5 attached
