Scientific software · Data engineering · Machine learning
Chicago, IL

Climate dashboard   /   Collider study explorer   /   LinkedIn   /   Email

I build scientific software and data systems, from calibration mathematics and database migrations to machine learning and interactive research applications. I studied Physics and Data Science at Lake Forest College, and I now split my time between hands-on laboratory work and developing software for laboratory operations.

My work follows the full path from a measurement or raw record to a result someone can inspect and use. That includes checking assumptions, validating calculations, designing data models, and explaining what a system does when something goes wrong.

Spatial statistics · Data engineering · Scientific communication

How do countries' warming rates align with geography, historical emissions, and development? I built an 8.6-million-row DuckDB pipeline covering 3,510 locations, with national comparisons, spatial validation, and a public Streamlit dashboard.

The completed model covers 151 countries. Geography accounts for the largest share of the observed variation, but the model predicts much less well when nearby countries are excluded from validation. A separate ERA5 product check showed that the spatial models' predictive gains depended on the temperature dataset. These checks shaped the conclusions.

Python DuckDB SciPy Plotly Streamlit
Explore the dashboard · Methods and findings

Scientific machine learning · PyTorch · Model evaluation

What information helps a model distinguish simulated particle jets? Building on earlier image-classification coursework, I carried out a separate 2026 study using ATLAS simulated jets, custom PyTorch models, and engineered physics measurements.

On 100,000 held-out simulated jets, adding physics measurements improved the compact ParticleNet model's mean unweighted ROC-AUC from 0.828 to 0.877 across three training runs. Boosted trees using those measurements reached 0.930. The comparison shows why the information supplied to a model matters. The public explorer explains the separate tasks and provides downloadable result tables.

Python PyTorch scikit-learn Altair Streamlit
Explore the study · Methods

As a Calibration & Radiochemistry Technical Lead, I work with gamma spectroscopy, alpha/beta counting, radiation-instrument calibration, and statistical QA/QC. Alongside the measurements, I develop the software and documentation that support them:

• Scientific calculation: shipped production Python calibration software with tested efficiency, detection-limit, and measurement-uncertainty calculations.
• Data engineering: migrated more than 6,000 instrument records to a normalized PostgreSQL database using Python ETL and post-load validation.
• Applications and reporting: delivered a client submission portal using TypeScript, Next.js, and MySQL, and deployed a Python tool that produces consistent client PDF and DOCX reports.
• Systems communication: translated inherited database logic and scientific calculations into workflow diagrams and a detailed technical guide for management and technical review.

• Numerical Methods: NumPy implementations of root finding, integration, ODE solvers, linear systems, optimization, and regression, with convergence checks and comparisons against established libraries.
• PDF Merge Utility: a practical reporting tool that pairs documents by filename identifiers, applies date filters, and merges PDFs in batches through a desktop interface.
• Sentiment Analysis: a modular Sentiment140 workflow with RoBERTa scoring, text preprocessing, batch inference, and visualization.

B.A. in Physics and Data Science · Lake Forest College
Magna cum laude · United World College alumnus