Hello! I'm Lawrence.

Feb 2026 — Jun 2026

Analyzed responses from a 2024 election survey of Black Americans that oversampled groups often underrepresented in polling, including Black men and Black Republicans. Used BERTopic to uncover themes in open-ended responses and HDBSCAN to identify voter subpopulations associated with socioeconomic factors such as income, employment, education, religion, and organizational membership. Used Splink to probabilistically link survey respondents with L2 voter records, enriching the survey data and assessing its accuracy.

Feb 2026 — Aug 2026

Developed and packaged a Python framework for training foundation models on mass spectra from small metabolites. Like protein language models that learn from amino acid sequences, these models learn spectral representations that can support tasks such as retrieving molecular structures directly from spectra. Built the training pipeline with PyTorch Lightning and established a maintainable software stack using Pixi, pytest, MkDocs, and GitHub Actions.

Dec 2023 — Jul 2025

Developed a framework for identifying and characterizing hadronic resonances through Bayesian inference over large parameter spaces, O(10⁶). The approach uses Gaussian processes and variational inference to model unknown components of mass spectra non-parametrically with phyiscal smoothness priors. These components can be interfered with parametric models of known processes, such as resonance line shapes, to describe the full spectrum.

Jul 2023 — Oct 2025

Built an end-to-end analysis pipeline for generating physics events, emulating detector efficiency with normalizing flows, and running rapid closure tests with Bayesian and frequentist inference. Added Python bindings for the collaboration's C++ analysis software and user-facing documentation to support long-term maintenance. Integrations with SciPy, Minuit, emcee, and NumPyro enable maximum-likelihood and Bayesian fits, making it easier to compare inference methods and diagnose method-specific pathologies.

Jan 2023 — May 2023

Published in The Astrophysical Journal: Collaborated with observational astronomers to develop a conditional variational autoencoder for generating template supernova spectra, reducing a class of systematic uncertainties by ~90%.

May 2021 — Aug 2022

Implemented a computationally intensive background subtraction technique in C++ with multiprocessing support, improving analysis efficiency for large datasets.

Jan 2021 — Oct 2021

Published in Physical Review D: Developed a framework using Deep Neural Networks to determine the nature of exotic hadrons from their spectra, utilizing Shapley values to understand feature importance.