Product Analyst
EXPERIMENTATION · BEHAVIORAL & OPERATIONAL ANALYTICS · DECISION SYSTEMS
I help teams turn behavioral, commercial, and operational data into decision systems by separating meaningful signal from noise before it becomes an expensive product, growth, or operating decision.
UW Foster MSBA · 3+ years analytics experience · MIT Sloan SSAC Runner-Up · Seattle
I specialize in separating meaningful signal from persuasive noise before a team makes a product, growth, commercial, or operating decision. My work spans B2B customer lifecycle analysis, marketplace and operations analytics, experimentation, and KPI systems. I have analyzed randomized experiments at a 576K-user scale, built risk and prioritization frameworks across customer and project portfolios, and reduced recurring decision latency from 48 to 15 hours by automating analytical reporting. I am most effective when the problem is ambiguous: a team has many possible signals, limited capacity, and an important choice to make. I define the outcome, examine plausible drivers, quantify uncertainty, and translate the analysis into a practical system, a score, a segment, an experiment readout, a KPI layer, or a prioritization rule that teams can use repeatedly.
Re-Systems Group Americas · Coon Rapids, MN
PC Connection, Inc. · Minneapolis, MN
Total Quality Logistics · Minneapolis, MN
University of Washington, Foster School of Business · GPA 3.85 (Top 10%)
User behavior → Experimentation → Causal rigor → Technical execution
Every number encodes an assumption about users. Name it before defending it.
The counterfactual comes first. The estimator follows from it, not the other way around.
Directional data gets a directional call. Only clean identification earns a firm one.
EDA and probability analysis of airline baggage delivery times across ~99K real flights.
EDA, feature engineering, and ZHVI price prediction across 1,406 West Coast cities over 10 years.
EDA and predictive response modeling for a direct mail campaign across 100K customer records.
Designed and analyzed a randomized controlled experiment across 576K users. Power calculations, balance checks, post-hoc segmentation for ITT and ATT effects. Detected heterogeneous treatment effects — inactive users showed 2.2% lift, non-American users 3.67%, while active/domestic users showed no effect.
Analyzed a randomized geo-experiment across 200 U.S. markets applying DiD with DMA and time fixed effects to isolate causal advertising lift vs. naive last-click attribution.
End-to-end ML system for football transfer market analysis: web scraping → S3 → Airflow → Databricks → MLflow → Kafka → Streamlit/Tableau.
Stock portfolio monitoring agent with automated alerts, technical indicators, and a self-serve Streamlit dashboard.
A personal agent-based decision-support system (Claude API + MCP stack) that scores and triages roles against a scraped JD corpus before I touch an application.