Things I've built — at the cross-section of data, finance & product.
End-to-end pipeline scoring loan applicants on alternative data signals — built to make underwriters' calls faster and more defensible.
Multi-touch attribution model for a SaaS product — built to answer the question 'which channel actually earned this dollar' in a way the marketing team trusted.
Side project — an automated rebalancing tool with tax-loss harvesting baked in. Started as a script for myself, ended up handling five asset classes and a small set of beta users.
A three-week funnel + qualitative research sprint for a digital bank. Mapped where users were dropping off, why, and what to do about it.
Scraped earnings call transcripts across eight quarters and built a sentiment model that quietly correlates with 30-day price movement. Now part of an internal research workflow.
Real-time flagging of unusual corporate-card transactions using unsupervised clustering. Designed to keep the operations review queue short and the false-positive rate honest.