Understanding people through data, building structure through technology.

Hi, I'm Shin Young Lee.

A Bridge Builder closing the gap between technology and people

Combining a foundation in Educational Technology with hands-on AI development,
I design solutions that put user experience and organizational growth at the center.

Profile photo

Education & Experience

Jun 2024 — Sep 2025

AI Service Developer & AI Instructor

Chunjae Education · Digital Learning Division

  • PoC and MVP development using Knowledge Tracing, GraphDB, and generative AI
  • Developed an API service for automated Korean college entrance exam (CSAT) Korean passage and question generation via GPT fine-tuning and LLM benchmarking
  • Built a math courseware MVP based on IRT and CAT
  • Data analysis and dashboard creation
  • AI lectures and mentoring

Related Projects

Genie
Feb 2025 – Sep 2025

AI-powered platform for automated CSAT Korean passage and question generation

Genie

Role  Planning & API Development

Problem

  • •High cost and time burden in creating passages and questions for exam prep workbooks

Approach

  • •Comparative evaluation of Claude, Gemini, GPT, and DeepSeek — blind testing with domain experts based on format (paragraph structure, sentence length, punctuation) and content quality
  • •Training data construction and GPT fine-tuning
  • •Prompt design based on 3-year CSAT question analysis; built an automated generation pipeline using a Gemini multi-agent system
  • •Iterative feedback cycles with the content team to refine generation quality

Outcome

  • •Reduced question creation and review time from several days to 1 hour
  • •Completed PoC; concluded before full production rollout
Python FastAPI Figma Fine-tuning Prompt Engineering RAG
CosMath
Nov 2024 – Jan 2025

AI-powered adaptive learning system for elementary school math

Role  IRT & CAT logic design and implementation, frontend development

Problem

  • •Existing service gave all students the same assessment, failing to account for individual learning levels
  • •Learning recommendations were provided, but students had no visibility into their own level or what prerequisite/follow-up topics to study

Approach

  • •Designed real-time difficulty estimation and next-item recommendation logic based on student response data (IRT-based ability estimation + CAT-based personalized recommendation)
  • •Visualized prerequisite and follow-up concept relationships as a knowledge map, giving students a clear view of where they are and what to study next

Outcome

  • •Delivered personalized learning recommendations and knowledge map visualization (PoC complete)
React TypeScript Tailwind CSS
Similarity-based Learning Chatbot Improvement
Aug 2024

Accuracy improvement for an in-house similarity-based learning chatbot for elementary students

Role  Planning & Development

Problem

  • •Low response accuracy in the existing keyword-based chatbot leading to poor engagement
  • •Generative AI could not be introduced due to hallucination concerns for a service targeting elementary students
  • •A gap existed between the business goal (increasing users) and the root cause (chatbot performance), with no one bridging them

Approach

  • •Analyzed Q&A data to identify the root cause; proposed adopting a keyword + embedding ensemble model
  • •Built and validated an open-source similarity ensemble model

Outcome

  • •Confirmed 20%+ accuracy improvement on frequently asked student questions; concluded at the proposal stage
Transformers Langchain FAISS

Aug 2025 — Oct 2025

Instructor

MetaCode

  • Instructional design and delivery for AI automation and data analysis courses

Jun 2022 — Feb 2023

Research Assistant

Korean Educational Development Institute (KEDI) · Educational Statistics Center

  • Published statistical report on school violence survey data
  • Event planning and coordination
  • Administrative support and document management
Publication

Feb 2019 — Feb 2024

B.A. in Educational Technology, Minor in Ethics

Ewha Womans University

Portfolio

Projects

LLMxiv
Jun 2025 – Sep 2025

Automated newsletter service that daily crawls, summarizes, and translates AI papers from arXiv Live

Role  Service planning, backend development, deployment & operations

Problem

  • •Hard to keep up with the volume of new AI papers published daily
  • •Wanted to build a fully automated service without paid servers or APIs

Approach

  • •Designed and automated an end-to-end data pipeline — unstructured text (paper abstracts) through collection, refinement, summarization, translation, and delivery
  • •Defined collection schema (same-day papers, category filters) and switched from RSS to a custom crawler based on source limitations to secure full abstracts
  • •GitHub Actions cron-based scheduling for zero-maintenance pipeline operation — no server cost
  • •Open-source Google Pegasus for summarization; DeepL API for Korean translation
  • •Google Sheets + GAS for subscriber management and sign-up UI; GitHub Pages for full archive

Outcome

  • •Delivers 5 LLM papers + 5 general AI papers daily, auto-summarized and translated — currently live
  • •Fully automated newsletter service with zero server or paid API costs
Python GitHub Actions BeautifulSoup Pegasus DeepL API Google Sheets GAS

Certifications

TOEIC SPEAKING (Level : Advanced Low)

Korea TOEIC Committee · Mar 2026

TEPS (Score : 354)

TEPS Council · Jun 2024

Google Analytics Certification

Google · 2026.04

Big Data Analytics Engineer

Korea Data Agency · Jul 2025

NCS Vocational Trainer
(IT Development, AI, IT Strategy & Planning)

Ministry of Employment and Labor · Jul 2025

AWS Certified AI Practitioner

AWS · Feb 2025

SQL Developer (SQLD)

Korea Data Agency · Jun 2024

Awards

Excellence Award — 6th Education Public Data Utilization Contest
2024.08

Korea Foundation for Science and Creativity · AI support service for students with special educational needs

Role  Service planning, data analysis, GPT fine-tuning & chatbot development

Problem

  • •The number of students requiring special education is rising sharply, while a shortage of trained teachers limits the support each student can receive
  • •Sentiment analysis of YouTube comments revealed that 62% of responses about inclusive education were negative, indicating a lack of social awareness
  • •No dedicated tool existed to provide individualized support for students with developmental disabilities who struggle with social skills such as emotional regulation and communication

Approach

  • •Fine-tuned GPT-3.5-Turbo on AI-Hub's Korean multi-session conversation corpus to build "Sangdami," a chatbot capable of empathetic counseling dialogue
  • •Designed system prompts to assign each chatbot a role tailored to the individual student's profile and behavioral characteristics
  • •Built a community feature open to students, families, and teachers — covering communication, information sharing, second-hand trading, and career resources
  • •Designed as an integrated platform supporting daily life across scheduling, journaling, games, and learning

Outcome

  • •Expected to improve social communication skills in students with special educational needs and reduce negative perceptions of inclusive education
Fine-tuning Prompt Engineering RAG

Skills

Overall

•

I design services that maximize user experience and outcomes grounded in instructional design principles, and I bring theoretical models to life as working products.

•

I don't assume the newest technology is always the right choice. I believe the best solution depends on context.

•

When a task demands it, I actively explore unfamiliar territory to deliver the best result. I've continuously expanded my own scope—from Python-based data analysis to LLM and GraphDB applications.

•

I enjoy pushing for improvements in team processes and culture. I'm especially drawn to experience design through HR data analysis.

Communication

•

I've frequently served as the bridge between product and engineering teams, translating requirements in a language both sides can understand.

•

Through AI and data analysis lectures and mentoring—including KDT(K-Digital Training) bootcamps and international programs—I've communicated technical concepts to learners from diverse backgrounds.

•

I speak up regardless of seniority or title, while staying genuinely open to being wrong.

•

I believe over-communicating is almost always better than under-communicating, and I prioritize sharing context to reduce friction in collaboration.

Languages

Python

End-to-end ownership: arXiv crawler, LLM pipeline, FastAPI server, and fine-tuning scripts—all built solo

SQL

Aggregate queries for learning data analysis, query optimization for dashboard integration

JavaScript

Workflow automation with Google Apps Script, basic frontend interaction

AI · Data

Transformers

Open-source model selection, application, and performance comparison; handwriting and image recognition model development

LLM

Training data construction for fine-tuning; local LLM serving and inference endpoint integration via Ollama; fine-tuning, RAG, and prompt engineering design and refinement; multi-LLM benchmarking

Knowledge Tracing

Adaptive assessment logic design and implementation based on IRT and CAT

Database

PostgreSQL

Schema design for service databases; aggregate query writing for learning history data analysis

Neo4j

Modeled prerequisite/follow-up math concept relationships as a knowledge graph; integrated with LLM-based GraphRAG pipeline

Backend

FastAPI

Sole designer and developer of LLM multi-agent REST API for the Genie project

Flask

Used for math chatbot backend prototyping

Frontend

React

Sole developer of the CosMath courseware frontend in TypeScript, including knowledge map visualization

HTML/CSS

Built responsive portfolio and service UIs using Tailwind CSS

DevOps · Infrastructure

GitHub Actions

Designed and currently operating a cron-based automation workflow for the LLMxiv project

Docker

Service containerization and separation of development and production environments

AWS (EC2/S3)

EC2-based API server deployment and operations; S3 static file management

Teaching

Audience Topic Format Notes
KDT PM Bootcamp AI, Machine Learning, Deep Learning, Personalized Learning Lecture
KDT Mentoring Adaptive learning courseware, step-by-step math chatbot, SEL chatbot — mentoring & development Mentoring
Seoul National University — Graduate students, Dept. of Educational Technology Building a teacher-assist chatbot with GraphDB & GraphRAG Seminar
Slides
Corporate staff, KDT participants & external attendees EdTech seminar on Knowledge Tracing & Recommendation Systems Seminar
Video
In-house PMs & Graduate students, Chinju National University of Education Vibe coding workshop using Google Apps Script & Gemini Workshop
MetaCode Students AI course: Generative AI, no-code automation, vibe coding, agent development Lecture Video
Kathmandu University, Nepal AI Project Mentoring Mentoring
Seminar photo