🇰🇷 KR · 🇬🇧 EN
Machine Learning Engineer | Decision Systems · Reinforcement Learning · Agentic AI
I build and ship ML systems that make decisions under real-world constraints — delayed feedback, changing data distributions, and hard compliance rules.
At Qraft Technologies, I own the ML pipeline for a PPO-based trade execution engine operating on live order book data across 3 equity markets, and the optimization core of a portfolio construction system serving retail banking customers. Previously, I applied ML to manufacturing time-series at LG Display.
I'm currently exploring agentic AI architectures for financial decision support and writing a research paper on RL scaling laws in sparse-reward regimes.
MEng Electronic & Information Engineering, Imperial College London
📍 Seoul, South Korea · Open to relocation
Links: GitHub · LinkedIn · Email · CV (PDF)
ㄴCore: Python · PyTorch · Rust (environment/inference integration)
ML/RL: PPO · GAE · CNN · LSTM · Transformers · LightGBM · SHAP · SciPy (optimization)
Agentic/LLM: LangChain / LangGraph · Pydantic · VectorDB · RAG · Ontology & Knowledge Graph · GraphRAG
Data: Pandas · NumPy · SQL · L2/L3 order book data · time-series feature engineering
Production: TorchScript · parallel training (16–128 workers) · batch pipelines · Rust ↔ Python IPC