CV
Education, research experience, teaching, honors, and selected projects.
Contact Information
| Name | Jeonghun Park |
| Professional Title | Integrated M.S./Ph.D. Student |
| top321902@snu.ac.kr | |
| Location | Seoul, South Korea |
| Website | https://pjh6029.github.io/ |
Professional Summary
Integrated M.S./Ph.D. student in Computer Science and Engineering at Seoul National University working on multimodal perception, commonsense reasoning, and agentic systems. I am interested in building reliable self-improving agents that connect perception, reasoning, and action in physical environments.
Experience
-
2025 - 2025 Seoul, South Korea
Research Intern
Human-Centered Computer System Lab, Seoul National University
- Proposed and implemented a Code-Text Action (CoTA) agent that unified textual reasoning with syntactic code execution and achieved a 1.6x accuracy gain over a ReAct baseline.
- Contributed to the Lumo project by developing AI agents and improving API-based tool-calling.
-
2024 - 2024 Seoul, South Korea
Software Engineering Intern
FADU
- Built a scalable RAG pipeline over company-wide technical documents using multi-level chunking and robust document parsing with the Upstage API.
- Implemented an abstraction layer for tool and provider integration, enabling flexible switching across multiple LLM providers and vector databases.
-
2022 - Seoul, South Korea
Teaching Assistant
Seoul National University
- Computer Programming (2022 Fall, Prof. Lee)
- Logic Design (2024 Spring, Prof. Lee)
- Programming Practice (2024 Fall, Prof. Jeon)
- Basic Computing (2024 Fall, Prof. Kim; 2025 Spring, Prof. Yamada)
Education
-
2026 - Seoul, South Korea
Integrated M.S./Ph.D.
Seoul National University
Computer Science and Engineering
- Physical & Interactive Intelligence Lab (Advisor: Youngjae Yu)
-
2020 - 2026 Seoul, South Korea
B.S.
Seoul National University
Computer Science and Engineering
- Major GPA: 4.09/4.3 (~4.28/4.5)
- Overall GPA: 3.93/4.3 (~4.13/4.5), summa cum laude
- 2022-2024 compulsory military service in the R.O.K. Air Force
Awards
-
2020 YongWoon Foundation Scholarship
YongWoon Foundation
Full undergraduate tuition scholarship spanning March 2020 through December 2026.
-
2025 Merit-based Scholarship
Seoul National University
Merit-based scholarship awarded in 2025.
Projects
-
Samsung NNTrainer Open-Source Contribution
Improved training efficiency in an open-source deep learning framework.
- Implemented gradient checkpointing, reducing training memory usage by up to 46%.
- Added Sophia and Lion optimizers to improve convergence speed and reduce memory overhead.
-
ARC-AGI Solver
Built an ARC-AGI solver for a deep learning course project and ranked 1st out of 35 teams on the hidden evaluation.
- Fine-tuned Qwen3-4B with LoRA and curated augmentation.
- Applied test-time training, grid-wise majority voting, and a fine-grained loss for efficient few-shot learning.
-
Minimal Deep Learning Framework
Implemented a PyTorch-like framework focused on educational clarity and efficient execution.
- Built a dynamic computation graph and define-by-run automatic differentiation engine from scratch.
- Added GPU support via CuPy.
-
Perspective-Switching RPG Game in Unity
Built an experimental game project around seamless 2D and 3D interaction.
- Led project architecture and team pipeline setup with GitHub, Slack APIs, and CI integration.
- Developed perspective-switching mechanics between 2D and 3D scenes.