Mathematics for the natural sciences

I build quantitative models for systems that don't hold still: crops, soils, watersheds, ecosystems. The work sits between physics and statistics, and its job is to turn field data into predictions a scientist or an agronomist can actually defend.

Background

I trained as a civil engineer, did my postdoc in agricultural science, taught as a professor, and now model for a government agency. Engineering taught me to take physical law seriously. The natural sciences taught me how rarely real systems obey it cleanly. Most of what I do lives in that gap, across numerical ecology, soil and agricultural modeling, hydrology, and geophysical earth science, with the occasional detour into health data.

Approach

I use whatever the problem actually needs: mechanistic models where the physics is known, statistics and machine learning where it isn't, and usually both at once. What I won't do is hand over a black box. A prediction is only useful if you can say how far to trust it, so calibrated uncertainty and validation that survives scrutiny are part of the work, not a footnote at the end.

What I do

I work with research institutions, industry, and government agencies to:

• Build predictive statistical and machine-learning models, and the decision tools around them
• Develop physics-based models with calibrated uncertainty
• Deliver reproducible computational environments that hold up in production
• Bring quantitative methods to agriculture, ecology, hydrology, and the earth sciences

The goal is the same every time: defensible answers from difficult data, in pipelines that still run a year from now.

Philosophy

Good science needs good tools. Complex systems need serious analysis, but analysis that can't be read and checked can't be trusted. I lean on reproducibility and automation to keep the mechanical parts honest and consistent, so that human judgment goes where it belongs: deciding what the results actually mean.