이건표 Gyunpyo Lee

Senior Data Scientist

LG Electronics

CX Center · BXI Team, Seoul, South Korea

Nov 2022 – Present

CLUE Insight Ecosystem (2023–Present)

• Led end-to-end AI product development from PoC to production deployment.
• Analyzed 26,000 respondents across South Korea, the U.S., and Vietnam, covering 3,500+ variables (purchase history, life events, customer preferences, demographics).
• Structured survey data into JSON-based non-relational database for scalable business intelligence.
• Built RFM clustering + LightGBM purchase prediction model achieving 83% accuracy (Precision 85%, Recall 80%, K-fold cross-validated).
• Developed LSTM life-event forecasting model with 87% accuracy predicting major life changes (education, job changes, relocations).
• Delivered data-driven insights for customer experience research and targeted marketing strategies.

Generative AI Implementation (2024–Present)

• Built Azure OpenAI embedding pipeline processing 26,000 × 3,500+ variables with FAISS vector indexing.
• Designed LangChain-based 5-agent system (Router + 4 specialized sub-agents) for automated data analysis.
• Conducted A/B testing revealing LLM hallucination issues in numerical computations. Made critical pivot decision: "Data analysis in Python, LLM as interface only."
• Rebuilt system using Azure Code Interpreter + Streamlit, successfully deploying service for 10-person team.
• Researched latest generative AI trends (GPT-4, Claude, RAG vs Fine-tuning tradeoffs, multimodal AI).
• Analyzed real-time sales data from 3 retail stores in Seoul to evaluate marketing effectiveness (cross-team collaboration).

AI Consultant

Code V

Seoul, South Korea

Mar 2022 – Dec 2024

Seafood Quality Inspection AI

• Led 6-month project managing data collection, model development, and deployment.
• Automated defect detection for three seafood categories (fishcake, seaweed, oysters).
• Coordinated with manufacturers and data collection vendors to acquire 5,000+ images from production facilities.
• Addressed long-tail problem through custom augmentation strategies (color transformation, texture variation).
• Fine-tuned YOLOv5 with hyperparameter optimization, achieving 98% precision and mAP 50 score of 87%.
• Optimized for edge deployment with model quantization and TensorRT conversion.
• Customized solutions for each category's production environment (lighting, surface reflection, irregular shapes).

Spectral Data AI

• Preprocessed spectral datasets and developed custom ANN models for seamless inference integration.

Chief Product Officer

Dream2Real

Daejeon, South Korea

Jul 2022 – Oct 2022

• Led synthetic data generation pipeline development for data-scarce industries (MTMC tracking, manufacturing defect inspection).
• Built digital twin environments using Unreal Engine & Blender for AI training data generation.
• Generated diverse scenarios (weather, lighting, camera angles, object movements) with automatic ground truth labeling.
• Conducted domain gap analysis between synthetic and real-world data distributions.
• Established AI data structuring methodology and 3D dataset validation processes.

Design Researcher

KAIST

Daejeon, South Korea

Aug 2020 – Aug 2022

Kia Design Center | PBV Vehicle Active Lighting Research

• Designed interior lighting scenarios using Unity + Projection Mapping for immersive experiences.
• Tested 50+ lighting pattern combinations (color temperature, brightness, patterns) for different use cases.
• Aligned scenarios with Kia's 'Opposite United' design philosophy through designer interviews.

LS Automotive | Affective Lighting Research

• Conducted EEG-based biometric testing with 40 participants (20 male, 20 female) to measure emotional responses to automotive lighting.
• Integrated quantitative biometric data (heart rate, skin conductance) with qualitative surveys.
• Validated findings using ANOVA and t-test statistical methods.

KAIST | Infection Control Hospital UX Research

• Designed wayfinding systems for patient/staff separation in infectious disease hospitals.
• Developed design guidelines through medical staff feedback and real-world testing.

Exterior Designer

Changan European Design Center

Turin, Italy

Feb 2016 – Dec 2019

• Key designer in Changan's first platform-based vehicle (UNI-V), leading exterior design refinement.
• Coordinated between CAD and Clay modeling teams using Autodesk Alias 3D CAD and Vred PBR rendering.
• Developed facelift models for production sedans (C301 / C201) based on competitive benchmarking (Geely, Dongfeng, Hyundai, Kia, Toyota, Honda).
• Designed Eado XT (C211) concept featured on Auto & Design 2018 cover.
• Contributed to Jiliu Concept and Yuyue Concept (Shanghai Auto Show 2017).
• Participated in brand strategy development and future autonomous vehicle UX research.