Led the development of EchoX, an LLM persona-based agentic system that automates end-to-end user research — from persona generation and autonomous task execution to behavioral data collection and analysis.

[My role in this project]
Leading the project end-to-end — designing the system architecture from persona generation through autonomous task execution to behavioral data collection and analysis.

Partnered with an university to measure the cognitive burden users experience in NAVER's search results, using eye-tracking, surveys, and interviews — translating findings into search experience design.

[My role in this project]
Leading the project — designing the mixed-methods study (eye-tracking, survey, interview), coordinating with University researchers, and translating findings into search experience design recommendations.

As AI-generated answers increasingly lead to zero-click search behavior, I researched how to evaluate user satisfaction with search experiences and results in this new environment, combining quantitative and qualitative methods.

[My role in this project]
Leading the project — designing the mixed-methods research framework to evaluate user satisfaction in a zero-click search environment.

Launching a virtual try-on service

Our web-based demo system to experience virtual try-on service

[My role in this project]
Conducting market research and interview, defining key features of an web application, creating user journey and interface layouts, developing an internal tool and a pipeline to create training and test dataset for the AI architecture.

By launching an AI avatar profile service, user acquisition rate has been increased by 92% and achieved 75% reach rate.

AI avatar profile images generated by our pipeline

[My role in this project]
Designing 900 prompt templates for figure and background, performing thematic analysis to quantify and interpret responses.

Creating a synthetic data generation pipeline for human pose and facial expression recognition models

Examples of synthetic training data generated by our pipeline

[My role in this project]
Interviewing AI engineers and animation graphic designers to understand the limitation of current data generation systems. Building training datasets worth $360,000

Understanding how people perceive 3D avatar facial expression features

Examples of synthetic training data generated by our pipeline

[My role in this project]
Designing and conducting a survey and focus group interviews to investigate characteristics of 3D avatar facial expression

Proposing qualitative evaluation metrics to assess human satisfaction on AI-generated music

Satisfaction with AI-generated music (1–7 Likert scale)

[My role in this project]
Leading the project

Designing a system that allow users to explore how AI architectures make prediction

Our system for analyzing how an attention-based deep learning model predicts traffic congestion

[My role in this project]
Executing some case studies to inverstigate user journey and identify the features of a visual analytics system

Suggesting a forecasting architecture under complex spatio-temporal condition

Multiple traffic data with similar pattern and the extracted representative pattern

[My role in this project]
Executing qualitative research to identify user pain points and behavior patterns

Investigating users' impulse buying behaviors and self-control strategies

Four types of interventions: (A) Reflection; (B) Distraction; (C) Desire Reduction; (D) Salient Cost

[My role in this project]
Conducting user surveys and interviews to identify user pain points and understand their behavior patterns. Investigating the experiment result by thematic analysis

Analyzing the performance of traffic forecasting architectures

Different models show varying accuracy performance based on traffic data dynamics

[My role in this project]
Investigating existing studies to select forecasting architectures and dataset for the evaluation. Writing a research paper.