One of our most complex technical challenges has been creating adaptive, engaging quiz systems that serve different learning objectives. Today, I'll break down how we built a multi-layered quiz generation system across our two platforms—each with distinct pedagogical approaches and technical implementations.
Main Platform Structured Learning Pathways
1. Topic & Subject-Based Quizzes
Topic-Based Practice
Students select specific topics → receive 15 targeted questions
Subject-Based Practice
Students select courses → receive 15 random questions across various topics within that subject
Study Mode
Non-pressure learning where answers with explanations are immediately visible alongside questions
2. JAMB Simulation
Authentic exam experience with optimizations:
- Reduced from 2 hours to 20 minutes per subject (focusing on speed development)
- Progressive difficulty curve matching official exam patterns
3. AI Playground
Student prompts AI with desired topic
"Generate 10 practice questions on Photosynthesis for JAMB level"
AI generates custom quiz questions in real-time
The Arena Gamified Progressive Learning
1. Level-Based Question Hierarchy
All questions categorized into 8 difficulty levels (1=easiest, 8=most difficult)
- Students must master current level to unlock next tier
- XP system tied directly to question difficulty and accuracy
2. Automated Question Classification Challenge
Problem: Manually categorizing 6,500+ questions by difficulty was impossible
Solution: Built a custom AI classification pipeline using OpenAI's API
- Process took days to complete
- Required handling API rate limits and costs
- Laptop crashes became routine (learning the importance of cloud processing)
- Simultaneously generated AI explanations and hints for each question
Significant investment in API costs but saved months of manual work
3. Brainstorm Mode: Interactive Word Challenges
Students guess missing letters in key terms:
Hint: "Basic unit of life"
- Progressive difficulty with increased missing letters
- Engages pattern recognition and concept reinforcement
💡 Key Learnings in Educational Quiz Design
🎯 One size doesn't fit all
Different learning objectives require different quiz structures
⚡ AI is an accelerator, not a replacement
Automated classification enabled scale but required careful validation
⏱️ Timing matters
Adjusting JAMB duration improved focus without sacrificing learning outcomes
🔮 Future Developments
- 🤝 Collaborative quiz modes for squad competitions
- 🔄 Real-time question generation based on current curriculum changes
📊 Technical Architecture Overview
// Simplified quiz generation flow
Quiz Types:
├── Main Platform
│ ├── Topic-based (15 questions, focused)
│ ├── Subject-based (15 questions, random)
│ ├── JAMB Simulation (160 total questions, timed)
│ └── AI Playground (dynamic generation)
└── Arena
├── Level-based (8 difficulty tiers)
├── AI Classification (6,500+ questions)
└── Brainstorm Mode (word completion)
// AI Classification Prompt (simplified)
const prompt = `
Categorize this question into difficulty level 1-8:
Question: ${question.text}
Subject: ${question.subject}
Topic: ${question.topic}
Consider: complexity, cognitive load,
and typical student performance.
`;🗣️ Question for You
Building educational technology requires balancing pedagogical effectiveness with technical feasibility. How are you approaching adaptive learning systems in your products?
💡 The Core Insight
The most effective quiz systems don't just test knowledge—they adapt to the learner's journey. By separating structured learning from gamified progression, we serve both the student who wants to master topics and the competitor who thrives on challenge.