AI-Driven Gamified Learning Platform (Bachelor's Thesis)
Adaptive Educational Platform with AI Integration – Backend Lead & Algorithm Developer
Leading backend development for an AI-powered adaptive learning platform targeting learners aged 7-18. Features intelligent assessment system, gamification elements, and adaptive difficulty algorithms based on educational research.
Technical Architecture
Built with modern Next.js App Router architecture, TypeScript for type safety, and Tailwind CSS with shadcn/ui components for responsive, accessible UI design.
Serverless backend using Next.js 14 API routes with TypeScript, handling authentication, assessment delivery, progress tracking, and adaptive difficulty adjustment.
Simplified 5-table schema (profiles, assessment_questions, user_assessments, user_assessment_answers, user_skill_levels) optimized for thesis scope, reducing complexity by 80% while maintaining core functionality.
Comprehensive RLS policies across 8 tables with 16+ rules ensuring COPPA-compliant privacy, user data isolation, and controlled social features. Automated database triggers for profile initialization.
Deployed on Vercel with automated CI/CD pipeline, serverless functions, and secure environment configuration for seamless production deployment.
Key Features & Implementation
- 36 multi-subject questions across Math, Reading, and Science
- 5-level skill proficiency determination (Beginner to Expert)
- Performance-based scoring with percentage thresholds
- Comprehensive tracking via user_assessments and user_assessment_answers tables
- Based on Zone of Proximal Development theory (Vygotsky, 1978)
- Knowledge Tracing implementation (Corbett & Anderson, 1994)
- Dynamic difficulty crossing: age-based levels overlap (e.g., Age 10 'hard' ≈ Age 11 'medium')
- Skill level adjustments based on assessment performance and user progress
- Points accumulation system integrated into user profiles
- Level progression tracking for motivation
- Streak tracking for consistent engagement
- Research-backed design to demonstrate impact on learning motivation
- Automated profile creation with handle_new_user() trigger
- Auto-initialization of adaptive learning state and progress tracking
- Email confirmation flow with callback routing and welcome experiences
- Database synchronization for seamless user onboarding
- OpenAI/Anthropic API integration for personalized recommendations
- Adaptive hints based on learner performance and context
- Difficulty suggestions with reasoning for transparency
- Learning pattern analysis identifying strengths and weaknesses
- ✓Reduced database complexity by 80% (from 11 tables to 5) while preserving core research functionality
- ✓Architected comprehensive RLS security framework with 16+ policies for COPPA compliance
- ✓Implemented automated database triggers for seamless user onboarding and initialization
- ✓Designed dynamic difficulty system maximizing question pool utilization across age groups
- ✓Built assessment system capable of precise 5-level skill determination across 3 subjects