👋🏻 My name is Jordan

🎓 Master of Information and Data Science @ UC Berkeley

🧩 I dig deep to find the connections between systems, people and patterns to build something that's measurably better

🔧 Proven experience working with teams to understand root causes and improve business processes

Forecasts demand across 1,800+ store-product series for a Latin American grocery chain, improving accuracy 27% over baseline and quantifying where better forecasts save the most money. Includes a disruption-resilience analysis around the 2016 Ecuador earthquake and an interactive Tableau dashboard.

Personalized product recommendations and item bundling are widely used strategies aimed at increasing consumer engagement and spending. This study uses a 3x2 factorial design to test the hypothesis that bundling and personalization strategies increase customer purchase amounts within a simulated e-commerce environment.

Modern large language models are trained on web-scraped data that often reflects various forms of bias. This bias can be exacerbated when models rely on it to make predictions, often resulting in negative effects towards certain groups. In this project, we explored the extent of gender bias in pre-trained modern LLMs like ModernBert, and applied 3 debiasing methods (counterfactual data augmentation, debiased embeddings, iterative nullspace projection) to reduce model bias in an occupation classification task related to resume screening workflows.

Utilizing open source data from the New York Metropolitan Museum of Art, our team converted the museum's collection data into a graph database, uncovering key insights into the relationships between objects and galleries that inform exhibition ideas, object curation, and user experience enhancements for visitor-facing products.

An exploratory data analysis project that uncovers relationships between gender, employment, and primary school enrollment at a global scale, producing insights into how early education relates to overall labor market trends.