Conducted research studying the particle nature of dark matter with large astrophysical datasets, resulting in three projects.Finding a Hidden Source of gamma-rays with Gaussian Processes (2023-2024) - Extracted a faint excess of gamma-rays from count data by training a Poisson model consisting of known background sources and a sparse variational gaussian process as the excess signal -- using NumPyro. JAX, and Tinygp. - Trained model on thousands of simulated datasets using over 100 GB of GPU memory in a GPU cluster to evaluate model performance and understand the effect of background mismodelling on the interpretation of the excess.Wavelet-Based Point Source Finder (2023-2024) - Developed an algorithm that applies the continuous wavelet transform on segments of the sphere for the detection of gamma-ray point sources across the entire sky, sped up by vectorized operations of NumPy and parallel computing. - Identified the optimal wavelet-based classifier by maximizing the AUC created by ROC curves formed using a range of thresholds for the wavelet values, achieving over 99% accuracy for point sources emitting more than 50 counts under realistic backgrounds. For fainter sources, accuracy depends on the brightness of surrounding backgrounds.Dark Matter Limits from Stellar Data (2021-2022) - Cleaned and processed the largest and highest-quality dataset of binary star systems} (1 million stars) using Pandas. - Developed a statistical model describing the evolution of the processed stars, their detection, and effects from dark matter. Distributions of measured stellar parameters and survival probabilities were empirically estimated with Monte Carlo simulations -- sped up by using over 600 CPUs for parallel computing. - Performed Bayesian inference with MCMC sampling to infer a posterior distribution, from which we were able to rule out possibilities for the mass and, for the first time, the shapes and sizes of dark matter in the Galaxy.

Taught live recitation sections for a physics course tailored to engineering majors. In addition to the duties of a recitation instructor, I worked closely with the lead professor to assist in course organization and management to deliver quality learning outcomes to students. Lead discussions on teaching plans involving the graduate teaching assistants and undergraduate learning assistants. Validated, obtained, and distributed course materials.Delivered recitation instruction for an online physics course tailored for engineering majors. Contributed to assessing and updating the structure, pace, and content of the course throughout the semester to ensure optimal student learning. Lead undergraduate learning assistants for clarifying instructions or solving problems requiring immediate resolution.

Standard deep learning sequence. Developed a more rigorous foundation of Deep Learning and acquired a formal introduction to PyTorch. I spearheaded the final project to completion: building and training a temporal graph neural network (GConvNN + LSTM) for predicting taxi demand that leverages both spatial and temporal features for making predictions, building and training a classical statistical model (ARIMAX) to act as a baseline model, analyzing and visualizing the raw data to eventually create a user-friendly pipeline for loading and processing NYC TLC taxi ride data, and creating a WebApp showing the changes in taxi demand over time.

Standard bootcamp giving me a formal introduction to regression, classification, data scraping, cleaning, and management. I received a certification by completing a project on analyzing the potential for company lobbying expenditures to be used as predictors for the changes in the price of their stock. We found the relation to be highly non-linear with no obvious trend, suggesting that a predictive model would be more effective if we implemented more features and restricted our list of companies, rather than analyze them all in aggregate.

Spearheaded a project studying celebrity deaths with statistical physics techniques. First, I developed a first-principles quantitative measure of fame based on the Bradley-Terry ranking model and measured it for 20 celebrities by conducting a survey of 50 undergraduate students. Then, I expanded our dataset by determining 4 internet-based predictors of fame and used them to infer that the idea that celebrities ``die in threes'' is a statistical consequence of a large number of famous deaths occurring each year.Created a model of bioluminescence in bacteria (Vibrio fischeri) to understand how chemical signals can facilitate population-wide changes in the bacteria. Grew bacterial samples, measured their bioluminescence, and fit the model to the measurements.Acted as lab technician in charge of diagnosing and repairing issues with existing equipment, requesting and ordering new equipment, and maintaining a lab encyclopedia on equipment usage.

Lead a project with the goal of measuring the optical properties of silicon and cuprate superconductors with Fourier-transform spectroscopy. Aided graduate students in sample preparation and low-temperature experimental setups. Performed maintenance for equipment in the lab.

Tutored students taking Physics 1/2 and Calculus 2. Met students through the Knack tutoring application, where students can shop around for their favorite tutor and request tutoring services at any time.

Part-time job at bowling alley. Primary responsibility was to organize, distribute, and sanitize bowling shoes. Additionally, I coordinated with other customer service representatives, front-desk employees, mechanics, bar-backs, and kitchen employees to ensure that customers were receiving satisfactory service, have their issues addressed, and have their bowling lanes fixed in a timely manner.