I'm a statistician with a broad focus on applied data analysis, statistical modeling, and interactive tools, spanning domains from clinical trial design and drug safety analysis to esports analytics. I hold a Master of Science in Statistics from the University of Illinois Urbana-Champaign (GPA 3.90), along with a Bachelor of Liberal Arts and Sciences in Statistics (GPA 3.77), with a minor in Mathematics, from the same university.
I work primarily in R and Python, with SAS as well, and have client-facing statistical consulting experience translating technical findings into stakeholder-ready recommendations.
Outside of my statistics work, I build games in Unity and C#, from solo long-term projects to 48-72 hour game jam entries. I enjoy building end-to-end projects generally, from data preparation and modeling to interactive tools and games.
You can find my statistics and data science projects below, followed by my game development work.
- Built a full CDISC SDTM/ADaM analysis pipeline in R on a public 254-patient Alzheimer's trial dataset, covering baseline demographics, adverse event summarization, and time-to-event survival analysis.
- Modeled time-to-adverse-event with Kaplan-Meier curves and a Cox proportional hazards model, uncovering a dose-response safety signal (HR of 4.15 and 5.03 vs. placebo) confirmed by an age-adjusted sensitivity analysis.
- Built a two-arm clinical trial sample size and power calculator in Python (Streamlit), covering continuous, binary, and time-to-event endpoints with closed-form formulas (noncentral t, arcsine, Schoenfeld) alongside simulation-based power that fits the actual regression model (ANCOVA, logistic regression, Cox PH) to synthetic trial data.
- Validated every closed-form formula against independent benchmarks, surfaced a real power gap between closed-form targets and Wald-test-based regression analyses across two endpoint types, and anchored a worked example to the HR of 5.03 finding from the CDISC safety analysis project.
- Ran a post-market drug safety signal detection analysis on FDA FAERS Q4 2025 data for semaglutide (Ozempic, Wegovy, Rybelsus), applying PRR, ROR, and chi-square disproportionality methods with Haldane-Anscombe correction across 3,279 primary-suspect case reports.
- Built the pipeline independently in Python and SAS, validated both implementations to match exactly, and curated the flagged signals against MedDRA coding noise and published GLP-1 safety literature.
- Analyzed CS:GO match and team-level data to study economy-driven decision-making across rounds.
- Merged and transformed match datasets to track transitions between eco, force, and full-buy rounds, and visualized their relationship to round outcomes using interactive plots.
- Developed a content-based recommender system to help indie developers identify potential publishers based on historical catalog similarity.
- Matched game concepts—defined by Steam tags and target price point—to publishers with comparable releases, and surfaces concrete examples to support discovery and outreach.
- Built text-based classification models to predict the geographic region of tweets using textual features. I focused on TF-IDF representations, class imbalance mitigation, and random forest modeling to improve predictive performance.
- The broader project also explored logistic regression baselines and BiLSTM models for sequential text patterns, implemented by other team members.