RL engineer from 🇰🇷 for now. I translate research-stage ideas
into deployable systems that hold up under real-world constraints —
most recently RL execution policies running live in APAC equity and
futures markets.
Led research and release decisions for live RL execution across APAC equity and futures markets, while extending the same evaluation discipline to trading-research agents and asset allocation.
Built and deployed RL systems for equity execution and futures trading, reducing live execution cost by 10bp+ and releasing internal strategies that sustained 100%+ annualized return at 3+ Sharpe over seven months.
Built CEX/DEX market-making infrastructure, including venue-aware simulators and low-latency on-chain data pipelines spanning 5+ global and 3+ Korean chains.
Supported the close of a ₩40B Series B and led the cross-functional launch of a triage-room waiting-time prediction service during the Omicron wave.
Replaced the Gaussian TD-error assumption with a learnable generalized-Gaussian model, improving uncertainty estimation and SAC/PPO sample efficiency under heavy-tailed dynamics.
Aligned N-BEATS feature distributions across source domains with Sinkhorn divergence, improving out-of-domain forecasting under severe distribution shift while preserving the model's interpretable basis decomposition.
Predicted continuous coordinate-MLPs from visible pixels with a transformer hypernetwork, improving reconstruction across unseen masks and OOD domains with roughly 7x fewer parameters than MAE-Large.
Personal Agents · Mechanical Watches · Golf & Tennis