Research

My research interests center on statistical machine learning, with particular emphasis on ranking and preference learning, graphical models and causal discovery, high-dimensional factor models, and network data analysis. I am also interested in the statistical evaluation of large language models and prediction-powered inference.

Statistical Machine Learning

I study statistical methods for learning from complex, structured, and high-dimensional data. My work is motivated by questions where uncertainty quantification, model structure, and computational feasibility all matter.

Ranking and Preference Learning

This direction concerns statistical models and estimation methods for rankings, comparisons, and preference data. I am interested in how to aggregate heterogeneous preferences and how to develop reliable inference procedures for ranking-related problems.

Graphical Models and Causal Discovery

I work on graphical representations of dependence and conditional structure, including models that help describe complex relationships among variables. Related questions include identifiability, estimation, and the statistical foundations of causal discovery.

High-Dimensional Factor Models

I am interested in factor-based methods for extracting latent structure from high-dimensional observations. These problems arise in settings where interpretable low-dimensional structure must be recovered from noisy and potentially dependent data.

Network Data Analysis

My research includes statistical modeling and inference for network data, including longitudinal networks, signed networks, and multilayer networks. A central goal is to understand latent structure, dependence, and uncertainty in relational data.

Large Language Model Evaluation

I am interested in statistical approaches to evaluating large language models. This includes developing principled ways to measure performance, compare systems, and account for uncertainty in human or model-based evaluation data.

Prediction-Powered Inference

I am also interested in prediction-powered inference, where predictions from modern machine learning systems are used to improve statistical inference while maintaining rigorous uncertainty quantification.