Research Orientation

My current research is shaped by two ongoing projects: one on data-driven AI for University of Michigan soccer and another on monitoring degradation in LLM coding-agent trajectories. Together they reflect my interest in decision support, optimization, and reliable applied AI systems.

Soccer analyticsAgent monitoringReliable LLM systemsOptimizationData-driven decision support

Ongoing projects 2
Application areas Soccer analytics + AI agents
Current emphasis Decision support, monitoring, applied ML

Current Ongoing Projects

At the moment, my research orientation is best described through two ongoing projects.

WinAI Propelling UM Soccer with Data-Driven AI: Mentor: Albert Berahas (aberahas@umich.edu). This project works with real collegiate performance, scouting, and tracking data to build analytical tools, predictive models, and decision-support workflows for the U-M Men's and Women's Soccer Teams.
LLM Agent Anomaly Detection: Mentor: Raed Al Kontar (alkontar@umich.edu). This project, carried out in collaboration with Rongbo Zhu, studies reliability questions in LLM coding agents. At a high level, it explores lightweight ways to monitor degradation in long agent trajectories before final failure.

Research Themes

Although these two projects live in very different domains, they share a common structure. Both ask how AI and data analysis can support decisions that matter in practice. One is embedded in sports performance and coaching, while the other is about making AI systems themselves more reliable and observable.

That combination fits the kind of problems I want to keep pursuing: settings where predictive performance, evaluation discipline, and clear decision relevance all matter at the same time.

Methods / Tools

My current toolkit is built around Python, statistics, and experiment-driven engineering workflows. Depending on the project, I move between predictive modeling, evaluation design, visualization, and structured technical reporting.

Python and R for data analysis and modeling
Statistical inference and predictive modeling
Optimization-oriented thinking for decision support
Visualization, reporting, and experimental comparison
Lightweight monitoring ideas for agent trajectories

Next-Term Goals

Next term, I plan to learn more through IOE 515 Stochastic Processes, EECS 498 Reinforcement Learning, and IOE 610 Linear Programming. I also want to keep exploring new fields while staying humble in front of knowledge.

Longer term, I hope to pursue a PhD.