A Survey on Leveraging Code Graphs for LLM-based Automated Program Repair
Shinyeob Kang, Jikwang Lee, Hyeokmin Kwon, Jaechang Nam. KCSE 2026 Proceedings, p. 170.
Undergraduate researcher working across AI interaction and software engineering.
Currently an intern researcher at the Ultimate Interface Lab, UNIST, studying AI interaction under Prof. Seongkook Heo.
I'm a Computer Science student at Handong Global University, working at the intersection of software engineering and human-computer interaction.
Most recently, I've been researching AI interaction as an undergraduate intern at the Ultimate Interface Lab, UNIST, advised by Prof. Seongkook Heo — looking at how people interact with AI agents, with a particular interest in accessibility.
Before that, my research centered on intelligent software engineering: LLM-based program repair, code graphs, and developer agents. I like taking a problem from its research question to a working system — designing the architecture, building the feedback loop, and learning from where it breaks.
B.S. in Computer Science, Advanced Major in AI Computer Science — Pohang, Republic of Korea
Relevant Coursework: Data Structures, Algorithms, Operating Systems, Database Systems, Machine Learning, Software Engineering, Compiler Theory
Finance with Accounting — Cape Town, South Africa
As an undergraduate research intern advised by Prof. Seongkook Heo, I study how people interact with AI systems, with a focus on accessibility — exploring interaction techniques and interfaces that make AI agents easier and more equitable to use.
As an undergraduate researcher in the Intelligent Software Engineering Lab, advised by Prof. Jae Chang Nam, I studied automated software engineering, code graph analysis, and AI/ML environment automation. I also contributed to static-analysis-based malware detection research in collaboration with UMV.
Shinyeob Kang, Jikwang Lee, Hyeokmin Kwon, Jaechang Nam. KCSE 2026 Proceedings, p. 170.
Hyeokmin Kwon, Jikwang Lee, Shinyeob Kang, Yongbean Chung, Jaechang Nam. KCSE 2026 Proceedings, p. 198.
An LLM-based agent that analyzes source imports, dependency files, and installation logs to generate and iteratively repair Conda environments.
Research into static-analysis features and benchmark pipelines for Linux ELF, Windows PE, PDF, Word, and Excel malware.
An MQTT delivery system with active/backup failover, QoS 1, store-and-forward, RTT-based routing, and duplicate filtering.
A LoRa sensing network with multi-hop relay, ACK-based delivery, CRC validation, and a real-time FastAPI dashboard.
Undergraduate Research Intern · Advised by Prof. Seongkook Heo · AI Interaction & Accessibility
Trainee, 13th Cohort
Research Assistant · Collaboration with UMV
Undergraduate Researcher · Handong Global University
Teaching Assistant · Handong Global University
Google AI Agent Program & Competition · Context Guard
Capstone Festival · Mitigating AI Hallucination
KCSE 2026 · Code Graphs for LLM-based APR Survey
Upstage AI Agent Hackathon · EnvAgent using n8n
Software Festival, Convergence Track · Vintan and Juum