Research Interests
Integer Programming
My research develops scalable methodologies for formulating and solving large-scale mixed-integer optimization problems. I investigate language-model-assisted mathematical modeling, learning-augmented optimization for accelerating search and decision processes, GPU-native algorithmic primitives, and decomposition-and-coordination methods such as surrogate Lagrangian relaxation. The overarching goal is to build robust end-to-end optimization systems that combine modeling intelligence, algorithmic structure, and modern computing architectures.
Large Language Model Systems
My research explores systems methodologies for deploying and serving large language models across heterogeneous computing platforms. It spans hardware-aware model placement, inference-runtime acceleration, memory and communication optimization, and automated end-to-end deployment. I am particularly interested in closed-loop systems that continuously profile execution, diagnose bottlenecks, and refine deployment and inference policies through iterative self-improvement.
Power Market Trading
My research studies algorithmic intelligence across the full lifecycle of power-market trading, from probabilistic forecasting and strategy construction to execution, risk-aware decision-making, and ex-post performance attribution. I develop AI-assisted decision frameworks and autonomous trading agents that integrate market information with physical power-system constraints, with an emphasis on transparent, adaptive, and operationally reliable decisions.