I like working on questions where the data is messy and no single source tells the whole story. Most of my projects sit somewhere between quantitative finance, alternative data, and applied statistics.
By alternative data, I mean using imperfect sources together—cash flows, sensor logs, weather records, evapotranspiration, fund data, or flood thresholds—and being clear about what they can and cannot tell us.
I'm building this portfolio for quantitative-analyst roles and data-science graduate programs.
My largest project. It compares private-market funds using dated cash flows, XIRR, investment multiples, scenarios, fee analysis, Monte Carlo simulation, and Excel checks. The point is not to predict returns. It is to show how assumptions about timing, fees, and terminal value change the decision.
A small Python ranking engine built from a larger fund-finder idea. It filters funds, scores rolling return and Sharpe ratio within peer groups, handles ties, and explains why a fund was excluded. The public version uses synthetic data.
My clearest alternative-data project. I combined evapotranspiration, NOAA weather, land-cover comparisons, and two very different climates to study golf-course water-demand signals. ET is a useful proxy here, but it is not the same as measured irrigation.
An R project covering regression diagnostics, transformations, model selection, seasonal decomposition, AR models, and forecast intervals.
- Environmental sensor quality control
- Flood-stage monitoring
- Climate, emissions, and space-data visualization
- Introductory Python projects
- Full project index
- Check the data before trusting the model.
- Use comparison groups and independent sources when one dataset is not enough.
- Keep assumptions visible.
- Prefer methods I can explain and test.
- Say when a result is exploratory instead of pretending it is causal.