Skip to content

HagXwon AI Learning Platform

HagXwon

Enterprise AI architecture for transforming Korea's $20B+ private education industry. An MBA + AI approach combining business strategy, technical architecture, and hands-on AI implementation.

Project Overview

Designed and prototyped a comprehensive AI-powered learning platform for Korean hagwons (private education centers) as part of the GenAI Bootcamp 2025. The project demonstrates the intersection of business strategy and AI engineering, addressing a $20B market opportunity with 70,000+ hagwons across South Korea.

MBA + AI Approach

This project showcases the unique combination of business acumen and technical execution:

  • Business Case Development: Market analysis, ROI projections, stakeholder impact assessment
  • Enterprise Architecture: TOGAF ADM methodology, C4 model diagrams, compliance frameworks
  • Technical Implementation: Built 6+ working AI prototypes using local models (Ollama)
  • Strategic Planning: 3-year phased implementation roadmap
Problem Statement

Despite $20B+ annual investment in hagwon education, students struggle with real-world English fluency. Traditional rote-learning methods prioritize test scores over conversational ability, leaving a critical gap in practical language skills.

Business Strategy

Market Opportunity

  • $20.6 billion USD spent annually on hagwon education in South Korea
  • 70,000+ hagwons operate across Korea, enrolling millions of students
  • 15M+ people worldwide actively learning Korean (7th most learned language globally)
  • English proficiency gap: Test-driven methods fail to build real-world fluency

Revenue Model

Model Description
B2B SaaS AI-powered tools licensed to hagwons
Direct-to-Consumer AI learning apps for students & parents
Freemium Model Free basic AI tools, premium features for a fee
Enterprise Consulting Custom AI integration for hagwon chains

Competitive Advantage

  • AI-Powered Personalization: Tailors lessons to individual learners
  • Scalability: AI tutors support thousands of students simultaneously
  • Regulatory Compliance: PIPA-aligned for secure student data management
  • Bilingual Focus: Korean-to-English AND English-to-Korean (not just one-way)

AI Applications Designed & Prototyped

Production-Ready Prototypes (Built with Ollama)

1. AI Speech Coach

  • Real-time pronunciation and fluency feedback
  • Accent analysis and correction suggestions
  • Progress tracking and personalized exercises
  • Tech: Local ASR models, speech recognition

2. Conversational Tutor

  • Interactive dialogue practice with AI personas
  • Context-aware conversation flows
  • Cultural and situational language learning
  • Tech: Ollama LLMs, multi-agent orchestration

3. Korean Sentence Constructor

  • AI-guided bilingual sentence formation
  • Grammar correction with explanations
  • Progressive difficulty adjustment
  • Tech: Local LLMs, structured output parsing

4. ASL Fingerspelling Recognition

  • Webcam-based Korean Sign Language (KSL) learning
  • Real-time fingerspelling detection
  • Interactive learning feedback
  • Tech: Computer vision, object detection models

5. Hangul Writing Evaluator

  • Automated assessment of Korean handwriting
  • Stroke order and character formation feedback
  • Comparative analysis with native patterns
  • Tech: Vision models, OCR, pattern recognition

6. Listening Comprehension System

  • Audio-based language learning exercises
  • Automated transcription and evaluation
  • Adaptive difficulty based on performance
  • Tech: Local ASR, audio processing

7. Multi-Agent Chatbot

  • Coordinated AI agents for different learning scenarios
  • Role-playing conversations (teacher, student, native speaker)
  • Context-aware dialogue management
  • Tech: Multi-agent frameworks, Ollama orchestration

8. Korean Learning MUD Game

  • Text-based adventure game for language learning
  • Interactive conversations with AI family members
  • Cultural context and vocabulary building
  • Tech: Game engine, conversational AI, CrewAI

Enterprise Features (Designed)

RAG-Powered Instructor Assistant

  • Knowledge retrieval for teachers
  • Lesson plan generation
  • Student progress insights
  • Curriculum optimization recommendations

Enterprise Architecture

TOGAF ADM Methodology

  • Preliminary Phase: Architecture principles and governance
  • Architecture Vision: Stakeholder requirements and business goals
  • Business Architecture: Process flows and organizational structure
  • Information Systems Architecture: Application and data architecture
  • Technology Architecture: Infrastructure and deployment strategy
  • Opportunities & Solutions: Implementation roadmap
  • Migration Planning: Phased rollout strategy

C4 Model Diagrams

  • Level 1 - Context: System boundaries and external dependencies
  • Level 2 - Container: High-level technology choices
  • Level 3 - Component: Internal structure of containers
  • Level 4 - Code: Class diagrams and implementation details

Zero Trust Security

  • Never trust, always verify
  • Least privilege access
  • Micro-segmentation
  • Continuous monitoring

Technology Stack

AI/ML Components

  • Local LLMs: Ollama for privacy-preserving AI
  • GenAI: GPT-4, Claude for cloud-based features
  • RAG: LlamaIndex, LangChain for knowledge retrieval
  • TTS: Text-to-speech for pronunciation modeling
  • ASR: Automatic speech recognition for evaluation
  • Computer Vision: Object detection, OCR for visual learning
  • Multi-Agent: CrewAI for agent orchestration

Infrastructure

  • Cloud: AWS (SageMaker, Lambda, S3)
  • Local Deployment: Ollama for on-premise AI
  • Containers: Docker, Kubernetes
  • CI/CD: GitHub Actions, Jenkins
  • Monitoring: Grafana, Prometheus
  • Database: PostgreSQL, Redis

Development

  • Languages: Python, TypeScript, React
  • Frameworks: FastAPI, Next.js, Tailwind CSS
  • Tools: Hugging Face, Weights & Biases, MLflow

Key Deliverables

Business & Strategy

  • Comprehensive business case analysis
  • Market research for Korean hagwon industry
  • ROI projections and cost-benefit analysis
  • Phased implementation roadmap (3-year plan)
  • Stakeholder impact assessment

Technical Architecture

  • Enterprise architecture using TOGAF ADM methodology
  • C4 model diagrams (Context, Container, Component, Code)
  • Microservices architecture design
  • Zero Trust security model
  • PIPA compliance framework (Korean privacy law)

Working Prototypes

  • 8 functional AI applications built with local models
  • Demonstrated feasibility of AI-powered learning
  • Validated technical approach with real implementations
  • Showcased privacy-preserving AI with Ollama

Compliance & Governance

  • Technical requirements matrix
  • Data protection and privacy controls
  • Bias mitigation strategies
  • Explainability framework for AI decisions
  • Accessibility standards (WCAG 2.1)

Implementation Roadmap

MVP (Weeks 1-6)

  • AI Speech Coach deployment
  • Sentence Constructor launch
  • Initial hagwon pilot program

Phase 2 (Weeks 7-12)

  • AI Live Conversations rollout
  • Homework Evaluator integration
  • Expanded pilot to 5+ hagwons

Phase 3 (Weeks 13+)

  • Full LMS integration
  • Enterprise features deployment
  • National hagwon network expansion

Impact & Value Proposition

For Students

  • 24/7 Practice: AI tutors available anytime
  • Personalized Learning: Adaptive difficulty and pacing
  • Safe Environment: Practice without fear of judgment
  • Immediate Feedback: Real-time correction and guidance
  • Real-World Fluency: Focus on conversational skills

For Teachers

  • Reduced Workload: Automated grading and feedback
  • Better Insights: Data-driven student progress tracking
  • Enhanced Lessons: AI-generated supplementary materials
  • Focus on Pedagogy: More time for high-value teaching

For Hagwons

  • Competitive Advantage: Modern, tech-enabled learning
  • Scalability: Serve more students without proportional cost increase
  • Quality Consistency: Standardized AI-powered instruction
  • Data-Driven Decisions: Analytics for curriculum optimization
  • Higher ROI: Improved student outcomes and retention

Skills Demonstrated

MBA + AI Integration: Business strategy, market analysis, financial modeling, technical architecture

Enterprise Architecture: TOGAF ADM, C4 modeling, system design, stakeholder management

AI/ML Engineering: RAG pipeline design, multimodal ML (speech, vision, text), prompt engineering, model selection

Local AI Deployment: Ollama orchestration, privacy-preserving AI, on-premise model management

Multi-Agent Systems: Agent coordination, role-based AI, conversational orchestration

Computer Vision: Object detection, OCR, image recognition, real-time processing

Compliance & Governance: PIPA (Korean privacy law), GDPR principles, bias mitigation, explainability frameworks

Research & Analysis: Academic dissertation, literature review, qualitative analysis, technical writing

Project Management: Roadmap planning, phased implementation, risk assessment, Agile delivery

Cross-Functional Collaboration: Business case development, technical documentation, stakeholder presentations

Research Contribution

Dissertation: Turn Detection in AI-Powered Language Learning

Conducted comprehensive analysis of conversational AI design for language education:

  • Technical Analysis: Turn-taking algorithms and latency optimization
  • Cultural Considerations: Korean communication patterns and politeness levels
  • Neurodivergent Design: Accommodating diverse learning styles and processing speeds
  • Ethical Implications: Privacy, bias, and accessibility in AI education
Key Finding

Effective AI language tutors must balance technical responsiveness with cultural appropriateness and cognitive accessibility.

← Back to Projects