Multi-wavelength Classification of Active and Star-forming Galaxies on the BPT Diagram with Supervised Machine Learning Models

Published in Publications of the Astronomical Society of the Pacific (PASP)

Presented at AAS 245 & 2025 APS Global Physics Summit; 2024 Princeton University Science Fair 1^st Place Overall.

Distinguishing between active galactic nuclei (AGN) and star-forming galaxies (SFGs) in large astronomical surveys is an important challenge with applications to various subfields of astrophysics and cosmology. Our study shows machine learning models applied to photometry can achieve over 90% accuracy, improving to 98% with combined spectroscopy.

Read full paper

Improved Fermi Blazar Candidate Classifications with SRG/eROSITA X-ray Counterparts using Machine Learning

Accepted for Publication in Publications of the Astronomical Society of the Pacific (PASP)

Presented at MIT IEEE Undergraduate Research Technology Conference; Columbia Junior Science Journal Semifinalist.

Using machine learning on gamma-ray and X-ray data from recent surveys, this study classifies active galactic nuclei with 99.2% accuracy, yielding better performance than traditional methods and models using gamma-ray data alone. The addition of X-ray data improves classification confidence and the number of sources identified. The model also predicts the types of 221 blazar candidates, demonstrating the strong potential of ML in multi-wavelength astronomical classification.

Read full paper

Repeating Flares, X-ray Outbursts and Delayed Infrared Emission: A Comprehensive Compilation of Optical Tidal Disruption Events – TDECat

Under Review for Publication in Astronomy & Astrophysics (A&A)

TDECat is a catalog of 134 confirmed tidal disruption events (TDEs) detected through 2024, including multi-wavelength photometry and publicly available spectra. Statistical analysis of flare timescales reveals log-normal distributions, and spectral classifications align with expectations from main-sequence star disruptions. Four new candidate repeating TDEs are identified, and a strong correlation is observed between infrared and X-ray emission. The findings indicate links between spectral class and multi-wavelength behavior, enabling robust population studies for upcoming surveys.

View preprint on arXiv

Hot Button Issue: Staying Cool as the World Heats Up

MathWorks Math Modeling Challenge (M3 Challenge) Finalist

This paper, selected as a Top 6 Finalist out of 794 entries in the 2025 M3 Challenge, models the effects of extreme heat on Memphis. It predicts indoor temperatures without AC, future peak energy demand, and neighborhood vulnerability using physical modeling, regression analysis, and entropy-based weighting. Results inform equitable resource allocation during heat crises.

View Paper

Surface of the Moon

2024 USA Young Physicists Tournament Champion

A 3D map of the Moon's near side was generated with the following methodology. The Lambertian photometric stereo algorithm was applied to images captured of the moon to compute surface normals, which were then integrated to reconstruct the lunar topography at kilometer-scale resolution.

View Project

A Scintillating Conversation

2025 USA Young Physicists Tournament Best Presenter Award (2x), 3^rd Place

This project examined atmospheric scintillation to explain why stars twinkle while planets do not. Intensity fluctuations were recorded using streak photography of Sirius, Betelgeuse, Jupiter, and Mars. Simulations modeled photon propagation through turbulent atmospheric layers using Planck’s law for blackbody spectra and phase screens generated via the DFT method. Results confirmed that stars, due to their small angular diameters, show significant scintillation, while larger planetary disks like Jupiter and Mars remain stable. Refraction and turbulence were identified as the primary causes.