Hi, I'm Gyuna π
I am an AI researcher and engineer with an M.S. from KAIST ICLab, advised by Prof. Uichin Lee. My work focuses on building and evaluating AI systems that learn from real-world, multimodal data, with experience spanning LLM evaluation, context engineering, agent systems, and sensor-based modeling. Across projects in healthcare, interactive systems, and applied AI, I have been interested in how data, models, and system design come together to support reliable and useful decision-making. Most recently, I studied how context engineering affects the performance and efficiency of LLMs in passive sensing-based inference. Ultimately, I aim to build AI systems that are practical, interpretable, and adaptable to real-world environments.
Education
Interests
AI/ML
Applications
News β¨
Publications π
COMPass: Literature Review and Systematic Evaluation of Context-Oriented Mental Health Modeling from Passive Sensing
Gyuna Kim, Youngji Koh and Uichin Lee
A Multimodal Sensor Fusion Approach Using Mobile, Wearable, and IoT Sensors for Mental Health Detection
Youngji Koh, Gyuna Kim, Chanhee Lee, Panyu Zhang, Yunhee Ku, Inhwan Choi, Jewoo Ryu and Uichin Lee
Submitted to IEEE Journal of Biomedical and Health Informatics,
Research Projects
Systematic Evaluation of Context Engineering for LLM-based Passive Sensing
Studying how context engineering choices affect the performance and efficiency of LLM-based inference from passive sensor data. I designed the evaluation pipeline and compared representation, retrieval, in-context learning, and reasoning strategies across multiple LLMs and datasets.
Multi-Agent Clinical Support System for Mental Health
Designed a multi-agent system that combines role-specific expert agents and a moderator to support mental health screening and lifestyle coaching. I built the sensor summarization pipeline, implemented the pilot system and dashboard.
Physical AI Safety for Industrial Environments
Analyzed safety risks in Physical AI, VLA, and agentic AI systems from the perspective of industrial safety management. I structured key risks such as jailbreaks and agentic misalignment, and contributed to proposal and paper drafting for safety-aware system design.
Multimodal Mood Detection with Mobile, Wearable, and Smart Home Sensors
Developed multimodal models for detecting depression and anxiety from mobile, wearable, speech, and smart home sensor data in collaboration with LG Electronics. I contributed to methodology, model development, and analysis.π» GitHub repository
Multimodal Stress Detection for Call Center Employees
Built multimodal models for detecting employee stress from Korean audio and text data collected in call center settings. I worked on data construction, utterance-level preprocessing, and language-model-based stress detection, and the project was supported by the NRF Graduate Research Fellowship for Master's Students.π» GitHub repository
Interactive Reporting System for Digital Health Data
Designed an interactive reporting system to help evaluators explore and interpret digital health data more clearly and flexibly. I led the dashboard design and analysis of user needs for presenting complex longitudinal health information.
Honors and Awards π
Experience
Work & Internship
Teaching & Academic Support
Teaching Assistant, CS565_DS522 IoT Data Science
KAIST
Student Supporter, Starmooc Lecture Video Production
UNIST
Mentorship & Outreach
Mentor, Research Internship
KAIST ICLab
Mentor, Explore@UNIST β High School Science & Tech Leadership Camp
UNIST
Mentor, Club-to-Club Entrepreneurship Mentorship for High School Students
UNIST
Photos πΈ
2025 | K-DS Conference in Daegu π