Hi, I’m Ian Pang

I’m completing my Ph.D. in Physics at Rutgers University (New High Energy Theory Center) under Prof. David Shih, with an anticipated graduation in June 2025. My research develops machine-learning methods for collider physics with emphasis on fast simulation, anomaly detection, foundation models, and open data. The goal is not only to make existing analyses faster and more efficient, but to broaden the scope of data-driven discovery.

Fast calorimeter simulation. I design flow-based generative models (normalizing flows) as surrogates for GEANT4 calorimeter showers. These models deliver large speedups while maintaining fidelity, and the same learned densities/latents double as calibration tools and as rich anomaly scores.

Unified ML frameworks. A recurring theme is showing that one architecture can serve multiple physics roles: (i) accelerate detector simulation, (ii) support unsupervised/semi-supervised anomaly detection, and (iii) enable calibration and likelihood-based inference.

Foundation models for HEP. Recently I’ve explored pretraining on large collider datasets with self-supervised objectives, then fine-tuning for downstream tasks. This yields consistent gains across classification, regression, and simulation-adjacent tasks and points toward scalable, general-purpose ML for HEP.

Open science. I’m committed to accessibility and reproducibility—for example, contributing to initiatives such as the Aspen Open Jets dataset and releasing code/baselines where possible.

See recent work below for papers and slides.