About
I'm currently a Senior Applied Machine Learning Engineer at Shopify, where I build next-generation search models for Shopify storefronts and the Shop app using fine-tuned LLMs. I also developed an LLM-as-a-judge evaluation system to automate search quality assessment and reduce manual labeling overhead.
Before Shopify, I spent eight years at Capital One, growing from Principal Data Scientist to Senior Manager. I led teams across pricing strategy, causal inference, and risk modeling — work that guided a $1.5B credit portfolio, contributed $70M in incremental revenue, and helped prevent over $100M in potential quarterly losses.
My roots are in materials science — I hold a PhD from The Ohio State University and spent four years at Lubrizol as a research chemist, where I built statistical models to predict chemical toxicity and developed image processing tools — early applied ML work before I had a name for it. That scientific foundation shapes how I approach ML: methodical, experiment-driven, and skeptical of results that seem too good.
I like to call myself a hybrid engi-scientist — equally at home reasoning about causal identification, tuning an LLM eval pipeline, and explaining results to a business stakeholder.
Experience
Senior Applied Machine Learning Engineer
Shopify — Remote
- Built LLM-as-a-judge evaluation system to automate search quality assessment, improving relevance and reducing manual labeling costs
- Building next-generation Shopify storefront and Shop search models using fine-tuned LLMs; leading experimentation, evaluation, and rollout
Senior Manager, Data Science
Capital One — McLean, VA
- Led team designing causal models for customer management and pricing, guiding a $1.5B credit portfolio
- Delivered LSTM-based risk model enabling interventions that reduced potential quarterly loss by $100M
- Built causal model for credit utilization prediction, expanding program reach by 20%
Manager, Data Science
Capital One — McLean, VA
- Developed pricing model contributing $70M in incremental revenue from credit line increases
- Rebuilt transaction fraud model preventing $4M in additional fraud
- Led challenger model development for Capital One's Main Street risk model
Principal Data Scientist
Capital One — McLean, VA
- Validated behavior, fraud, and recommendation models within the Model Risk Office
- Built AWS-based data engineering pipelines for model validation workflows
Research Chemist
Lubrizol — Wickliffe, OH
- Built statistical model to predict chemical toxicity, avoiding $1M in regulatory registration costs
- Applied Design of Experiments to accelerate new product development and developed rheology modifiers
Skills
Languages & Libraries
Infrastructure
ML & Statistics
Open Source Projects
Smart Underwriter
SVM model trained on Fannie Mae single-family data (2000–2015) to simulate automated underwriting system credit decisions.
COVID-19 Testing in US
Visualization app showing testing volume and positivity rates by state, pulling live data from the Covid Tracking Project.
GPC Fit
Open-source peak deconvolution tool for Gel Permeation Chromatography data. Helps polymer scientists separate overlapping peaks.
Education
PhD, Materials Science & Engineering
The Ohio State University, Columbus, OH
MS, Materials Science & Engineering
Clemson University, Clemson, SC
BS, Macromolecular Science & Engineering
Fudan University, Shanghai, China
Full publication list on Google Scholar.