• Lead the migration of a business-critical modeling codebase from on-premises SAS to Python in the cloud (AWS). Technical requirements include the ability to process over 100 million auto policies representing billions of dollars in revenue, runtime performance that meets or exceeds the current solution, ease of use for less technical users, and thorough testing
• Use two-stage generalized linear models to appropriately price auto insurance policies worth over $10 billion of written premium
• Trained a propensity model in Python using XGboost (a gradient-boosting library) and used a matching algorithm in R to understand the change in customer profile after the introduction of a new insurance product
• Built a model to identify cross-selling opportunities as part of a data science competition. Out of a field of nearly 30 teams, came in second place.

• Added functionality and improved performance (in some cases over 2X) and maintainability of an Oracle back-end that powers a CRM application serving dozens of non-profit clients with a combined donor base in the millions
• Developed a model to improve the performance of a client’s direct mail campaign, increasing response rates over 20%
• Worked with client-facing teams to address issues, identify bugs, and better understand customers’ needs

• Conducted analyses and created visualizations using R, Python, SQL, and Tableau in projects related to climate change, hydraulic fracturing, drinking water contaminants, greenhouse gas emissions, and home energy use
• Developed database applications in Microsoft Access in support of a variety of EPA programs
• Created a web application in a highly secured, locked-down environment for the Army Corps of Engineers to prioritize assets most vulnerable to climate change

Certificate in Statistical and Computational Data Science. Coursework included regression, Bayesian statistics, machine learning, neural networks, natural language processing (NLP), survival analysis, design of experiments, and visualization

General Honors, Departmental Honors, Phi Beta Kappa Society. Publications in ACS Omega and Biophysical Journal

• Created a user-friendly Streamlit application to classify photos of dogs into one of 150 AKC-recognized breeds
• Scraped approximately 20,000 images of dogs from the internet for training and scoring
• Used fastai and transfer learning to fine-tune an ImageNet-based convolutional neural network (CNN), achieving a test set accuracy of 53.8%

• Developed a high-level API to download public comments from Regulations.gov
• Used Python to greatly simplify the download process by handling the complex pagination scheme, API request limits, and layers of requests needed to access comment data