Computer engineering student studying at Hongik University (B.E. expected Feb 2027).
An LLM-ensemble system that detects and classifies malicious packets.
Multiple models vote on what an attack is, and a knowledge graph verifies their answers.
Language models handle perception; the graph handles truth. Code is private due to a lab-industry collaboration. LangGraph Neo4j FastAPI
What can you learn just by listening to a machine? This project
reproduces an acoustic side-channel attack paper, then goes a few steps
further: better feature extraction and model tuning to push accuracy
beyond the original results. PyTorch Scikit-learn
→ presentation (video)
Google Earth, but for codebases. Zoom out to see the whole repository, zoom in to a single function. Built on tree-sitter parsing, with AST-based complexity metrics and graph-based reading order suggestions. Capstone project, in progress. → repo
News, served like a snack. Summarizes articles, explains difficult terms, and turns what you read into quizzes so it actually sticks. Built with a university developer club. I developed the user-facing frontend and gathered feedback through live demos. React Gemini API
→ repo
- ML pipelines: making them fast, testable, and reproducible
- Anomaly detection in security and log data
- Data tooling and code analysis