Data Analyst
I ship forecasts, pipelines, and analyses that hold up under questioning.
I'm a data analyst at Choice Hotels, where I build demand forecasts and revenue analytics for 80 properties. Before that I built claims-volume forecasting and ETL pipelines at Prime Therapeutics and worked on real-time fraud detection at Bank of America.
The through line in my work is rigor you can check: causal estimates validated with placebo tests, forecasts scored on MAPE against the model they replaced, and detection systems that grade their own precision and recall. If a number cannot survive questioning, I do not ship it.
I'm looking for data scientist, data engineer, and analyst roles on teams that hold their analyses to that standard.
A causal study of mid-season managerial changes across the top 20 European leagues over six seasons (2019/20 to 2024/25). Combined propensity score matching with difference-in-differences over a 68,404-match panel covering 2,053 firings.
Finding: firing causes a +0.292 xGD gain per match over the following 12 matchweeks (95% CI [0.192, 0.392], p < 0.001).
Built the full pipeline: API-Football and Transfermarkt scraping via Selenium into a 6-table SQLite store. Validated with covariate balance checks, event-study pre-trends, and placebo tests.
A daily data pipeline that collects from three public APIs (Hacker News via Algolia, Stack Exchange, Adzuna) into SQLite and applies rule-based, direction-aware scoring to classify posts as competitive-displacement signals. Includes false-positive controls (skills-list dampener, staffing-firm flags, stub containment) and honest empty-state handling.
A deterministic detection system over seeded synthetic SaaS usage data. Five configurable rule-based detectors cover consumption ramp, seat expansion, feature adoption, workspace growth with a persistence gate, and churn risk. A built-in evaluation harness scores precision and recall against hidden account archetypes, including nine boundary-case confuser accounts, and a multi-seed benchmark command runs threshold sweeps. Reproducible in a clean-room clone.
Compared logistic regression, decision trees, SVMs, and a PyTorch neural network on the UCI Adult dataset. Reached 91% validation accuracy through structured hyperparameter tuning, with cross-validation and model selection managing bias-variance tradeoffs for consistent holdout performance.
A real-time computer vision system that converts facial movements into game controls using MediaPipe facial-landmark tracking. Frame-processing smoothing is tuned for stable, low-latency input.
Jun 2025 to present
Business and data analystLos Angeles
Built a demand-forecasting model on two years of booking and ADR data that cut forecast MAPE from 18% to 12% across 80 properties and feeds the weekly revenue-management committee. Shipped four Power BI dashboards (occupancy, ADR, RevPAR, channel mix) on a star-schema model that replaced a manual Excel rollup. Designed an A/B test framework for three promotional campaigns; the winning treatment lifted loyalty engagement 7% over control. Segment-targeting work with marketing showed +15% conversion over the broad-list baseline.
Sep 2024 to Jun 2025
Junior data scientistRemote
Built an XGBoost claims-volume forecast that cut MAPE by 10% versus the prior linear baseline, consumed by Operations and Finance for capacity planning. Authored six Python and SQL ETL jobs ingesting claims, formulary, and member data from five source systems onto a scheduled, monitored pipeline. Feature engineering (25+ features) lifted model AUC from 0.78 to 0.84 on holdout.
Jun 2023 to Aug 2023
Data science internNew York
Co-developed a real-time fraud-detection pipeline on 2M+ daily card transactions, cutting the false-positive rate about 10%. Engineered transaction-velocity and merchant-cluster features that raised recall 9% on the weekly evaluation set. Refactored four SQL extraction queries for a 14% average runtime cut.
SQL (PostgreSQL, MySQL, T-SQL, window functions, CTEs, query optimization), Python (Pandas, NumPy, Statsmodels, Scikit-learn), and R. Power BI (DAX, Power Query, RLS) and Tableau. A/B testing and causal inference, time series. Snowflake, BigQuery, Databricks, dbt, AWS S3, Git.