Hi, I’m Philip

Designed and evaluated multiple models for automated scanning in microwave impedance microscopy (MIM) with methods including particle filter, CNN-based, and multimodal LLM approaches. Improved directional accuracy by 40% and reduced manual scanning time by 1-4 days using the particle filter method. Also helped create a hypercontrol GUI for experiments.

Analyzed luminosity datasets (SDSS, Fermi-LAT, VLBI) using non-parametric statistical methods in Python (numpy, scipy, astropy, pandas) to help understand accretion disk-jet correlations and their cosmological evolutions. Derived the radio-gamma luminosity funtion and applied Kendall’s tau and Pearson correlation tests, showing higher gamma-radio local luminosity correlations in comparison to radio-optical.

Tested semi-supervised methods for detecting beyond-the-standard-model tracking signatures by evaluating a proprietary permutation-invariant anomaly detection model (sci-kit learn), achieving results to the supervise Particle Flow Network. Also investigated oversampling artifacts in high-energy physics pile-up simulation data that introduced bias into model training.