A model is a claim about how something behaves. The question is whether it survives contact with the people who act on it.

I provide independent quantitative analysis to energy infrastructure and finance teams, working from a background in economics and roughly two decades of running operations where errors are immediate and public.

The economics came first: an M.A. in Finance and Banking and a B.A. in Economics from Charles University in Prague, both with honors, followed by work as an analyst at Erste Group on valuations and a real estate portfolio tender that closed at $160 million.

The operational years came next, at CNN, NBCUniversal, and Czech Television. I launched a business news channel on a $20 million investment, ran augmented reality graphics for an election night broadcast averaging over 18 million viewers, anchored more than 2,000 hours of live national news, and audited CNN's production operation from its accounting data. Live broadcast is an unusual place to learn systems design and a clarifying one. Every workflow either holds under load or fails in front of an audience, which teaches you to care about the failure mode before the feature.

That is what I bring to current work. The PJM grid model processes close to a gigabyte of hourly market, load, interconnection, and geospatial data to rank regional vulnerability, with three methods that answer three different objections. Blackletter interprets legal documents where a confident wrong answer is worse than no answer. Both are built so the reasoning can be inspected rather than trusted, and both are running and open to inspection.