Research

My work bridges statistics and machine learning, with a focus on developing principled methods that make data a more powerful storyteller. I am particularly interested in problems where classical statistical thinking can meaningfully improve modern AI systems.

🔀
Mixture Reduction
Approximating large or growing mixture models with simpler ones, with applications to federated learning and 3D Gaussian Splatting compaction.
Read more →
📊
Empirical Likelihood in Modern AI
Modernizing classical empirical likelihood for federated learning and noisy label settings, shifting FL from aggregation to intelligent guidance.
Read more →
🗂️
Tabular Foundation Models
Evaluating and extending prior-fitted networks (PFNs) for in-context statistical learning — from supervised prediction to clustering.
Read more →