MSE Data Science · Los Angeles
I build pipelines, models and tools that turn raw, messy data into something people can decide with.
I'm a data scientist by training with a background that runs from computational linguistics to database systems. I like the whole path from raw source to finished answer: pulling the data, modeling it so it stays correct, and building the report, dashboard or automation that someone actually uses.
Most of my work sits where analytics meets operations. I care about getting the grain of the data right, checking numbers against known totals, and saying clearly what a result does and doesn't show.
Can a model built only on public data forecast California's power grid as well as the grid operator does? Every morning Grid forecasts the next day's demand, solar and wind output hour by hour, locks the forecast in, and grades it against what actually happened and against CAISO's own forecast. On 60 held out days the solar forecast matched CAISO's (9.7% vs 9.8% error). Behind it, Python jobs load CAISO and weather data into PostgreSQL every 5 minutes, and the dashboard also tracks curtailment, negative prices and the cleanest hours to charge an EV.
A decision tree regressor written from scratch in NumPy and trained on German used car listings. It matches scikit-learn's accuracy at the same depth (R² 0.75).
A two step MapReduce job that merges two booking datasets with different schemas and splits multi month stays across months to get revenue by month.
An Excel data model in Power Pivot with lookups, eight PivotTables with slicers and a VBA lookup macro, built on world population and language data.
A peer reviewed corpus study of how Russian speaking Ukrainian vloggers shifted toward Ukrainian after the full scale invasion. I sourced the video, annotated more than 10 hours of footage from three vloggers and built the visualizations. The shift showed up over months, where language change usually takes generations.
Read the paper →SQL (PostgreSQL), Python, pandas, NumPy, scikit-learn, C++
Excel (Power Pivot, VBA), Power BI, Plotly Dash, dashboards
ETL, data modeling, forecasting, statistics, machine learning, automation