Seonghyun

I am Seonghyun — a full-stack and machine-learning engineer working on agentic AI systems.

My current focus is the design and evaluation of AI-agent architectures: orchestrator–worker loops, tool-augmented reasoning, and the conditions under which delegating decisions to an AI helps or quietly accumulates risk. On the modeling side I build reinforcement-learning and stochastic risk models of human–AI interaction — when an agent should consult a model, how trust and bias migrate over repeated interactions, and how to keep a long-running agent loop honest.

This site collects my notes, working papers, and a research blog. The posts lean technical and equation-heavy; they are written for readers who want the math, not just the intuition.

Profile photo (assets/img/prof_pic.jpg) is still the theme’s placeholder image — pending a real one.