About

I am an incoming M.S. student at KAIST FlexML Lab, advised by Prof. Insu Han. I joined the lab in September 2026 and will begin the M.S. program in Spring 2027. I am currently finishing my B.S. in Computer Science at KAIST.

My research interests are AI agents and the decision models behind them. Much of an agent's work is not writing but deciding: which tool to call, where to route a request, whether a step succeeded. I am interested in models built for these decisions, such as Jev, which return typed choices with calibrated confidence instead of free-form text.

Before joining FlexML, I studied efficient attention and LLM agents at the SIIT Lab under Prof. Junmo Kim and spent a semester at the Technical University of Denmark studying computer vision. Outside the lab, I built LLM-based services as an intern at NewXphere and, before that, interned at Penta Security, where I saw firsthand how web security and cryptography research is done.

Contact

Email: 1000luca@kaist.ac.kr
Daejeon, South Korea

News

  • Joined KAIST FlexML Lab, advised by Prof. Insu Han. I will begin the M.S. program in Spring 2027.
  • Completed the LG AI Aimers 7th program on menu demand forecasting.
  • Started an AI service development internship at NewXphere.
  • Began an exchange semester at the Technical University of Denmark.

Research experience

FlexML Lab, KAIST

Sep 2026 – present

Undergraduate researcher (advisor: Prof. Insu Han)

Efficient attention and LLM agents

2025 – Aug 2026

Independent study, SIIT Lab, KAIST (advisor: Prof. Junmo Kim)

Studied RoPE-based positional encodings and linear attention, comparing their long-context behavior and compute–memory trade-offs with standard softmax attention. Surveyed LLM agents with a focus on memory, planning, tool use, reasoning, and test-time computation.

Cost and score metrics for object detection

2024

Course research project, CS470 Introduction to Artificial Intelligence, KAIST

Studied evaluation metrics for object detectors, including mAP and OC-cost. Proposed a modified OC-cost formulation and compared it with the original metric across multiple models.

High-efficiency wet air purification system

2020 – 2021

Korea Science Academy of KAIST, with the KAIST Civil and Environmental Engineering Lab

Developed and operated a high-efficiency wet air purification device. The work won the Gold Prize at the 18th Korea Science and Engineering Fair and was selected to represent Korea at Regeneron ISEF.

Education

KAIST

From 2027

M.S. student, FlexML Lab (advisor: Prof. Insu Han)

KAIST

2022 – present

B.S. in Computer Science

Technical University of Denmark

Feb – Jun 2025

Exchange student, computer vision and AI coursework

Korea Science Academy of KAIST

2019 – 2022

High school diploma

Experience

NewXphere

Jul – Dec 2025

AI service development intern

Built a RAG-based chatbot that recommends careers to elementary school students and an AI moderator for discussion-based training, and contributed to an AI employee education service developed with CJ CheilJedang.

Penta Security Systems

Jul – Aug 2023

Intern, Security Technology Research Center, Crypto Development Department

Gained hands-on exposure to enterprise web security and cryptographic technology research, and built a foundation in secure system design.

Honors and programs

  • Google for Startups AI Startup School 2025, completed
  • LG AI Aimers 7th, completed
  • Gold Prize, 18th Korea Science and Engineering Fair (KSEF)
  • Ricoh Sustainable Development Award
  • Korea representative, Regeneron ISEF