Fraud modeling and risk mitigation decision design for digital payment systems (Zelle, TransferNow, etc)

Pioneered and developed data science and data analytics solutions, empowering Sandian scientists with data-driven decision-making capabilities in nuclear physics experiment operations. (Python, SQL, PowerBI)• Conceived, designed, and executed an end-to-end interpretable machine learning solution that predicts and analyzes operational failures, achieving 3x better precision than random chance, with a recall rate exceeding 50%.• Employed interpretable algorithms, including SHAP feature importance values, to provide transparent and actionable insights into model behavior and communicated these insights to stakeholders through PowerBI reports.• My work resulted in the creation of an operational analytics division, and I actively documented and presented project specifics to facilitate its replication in other organizations

• Spearheaded the establishment and management of a robust business intelligence data infrastructure, streamlining the ingestion and processing of raw data from various data sources using database connections and REST API endpoints.• Delivered data analysis and business intelligence insights to stakeholders through PowerBi reports and data visualizations with automated report updates through Gitlab runners and triggers.• Demonstrated strong documentation and communications skills by creating technical procedures and advocating for a data-driven culture, supported by a published report and presentations given to laboratory directors.• Technologies: Python, pandas, numpy, sql, Microsoft Azure, Power Bi, yaml, Git, Bash.

• Developed an interpretable classification solution to use customer marketing engagement behavior to prioritize potential sales opportunities at the enterprise level and understand marketing efficacy.• Successfully segmented organizations that were 1.5x, 2x, and 2.5x more likely to purchase across different firmographics, enabling sellers to be more strategic in their engagement strategies and to reduce their time spent sourcing leads.• Interpreted model behavior to inform future B2B marketing strategy and to analyze the efficacy of previous marketing tactics.• Routinely presented to technical, executive, and non-technical audiences• Technologies: Python, SQL

• Lead the implementation of a novel natural language processing (NLP) and network analysis pipeline designed to unravel insights from thousands of pages of digitized diaries to perform named entity recognition (NER).• Performed custom label design, data labeling, and model training using the SpaCy language model to extract domain- specific information and enhance the NER analysis and performance (~10% accuracy lift).• Showcased expert domain knowledge and effective communication abilities through first authorship of a journal paper and leading international conference presentations attended by academic leaders in the domain space.• Technologies: Python, pandas, spaCy, scikit-learn, Git, Gephi, Tableau.