- Evaluated machine learning algorithms in Python (pandas, scikit-learn) for generating synthetic household vehicle preferences for use in a policy simulation project used by several East Coast states on a contract worth \$130,000.- Utilized R and Arrow to build data cleaning and machine learning pipelines on 100+ GB datasets to analyze the economic impacts of vehicle taxation changes in California on a contract worth over \$115,000. - Translated economic models from Excel to MySQL and integrated them into our economic impact forecasting software so they could be used for transit planning in the state of Michigan on a contract worth over \$500,000.- Rewrote an Excel/VBA economic modelling tool in R, reducing its setup and runtime from hours to minutes while adding new features for handling data skew.

- Developed a unique research dataset utilizing Python and JavaScript scraper, amassing 80GB of broadcast news transcripts into a PostgreSQL database and identifying relevant entries with BERT-based topic modelling.

- Constructed data pipelines with Python (pandas) and SQL to download and clean 11.5 million media documents including textbooks and educational videos, transcribe audio and video with ffmpeg and Google Text-to-Speech, and store all resultant text in BigQuery.- Prototyped an NLP-based search and recommendation engine using spacy word2vec to create context-aware embeddings for textbook sections, deployed behind a Flask API with React frontend.

- Developed a custom Java ETL tool for automating data conversion and QA processes, saving an estimated 50 hours of business analyst and managerial time while eliminating manual errors.- Contributed frontend web accessibility fixes to a major data submission portal, avoiding over \$1.5M in fines.

- Developed React/Redux frontends and Node (Express) backends for nine educational simulations, resulting in over \$990,000 in revenue for clients including Harvard, MIT, and the CDC.- Created a internal Node CLI tool to generate project boilerplate, significantly reducing developer time per project and enhancing compliance with internal coding standards and saving approximately 80 hours of developer time per year for six years and counting.

- Improved analysis of 11,000 points of string data by visualizing data in Python.- Provided editorial support for quantitative PowerPoint reports worth over $52,000 in revenue.