Path Dependence Research

Industrial forecasting at export scale.

Production engagement, anonymized · ~20 liquefaction trains · forecast error under 0.5%

For one of the largest U.S. LNG exporters: production forecasting across roughly twenty liquefaction trains, with forecast error under half a percent — accuracy that informed planning across the operation.

Simulation before it had a name.

Federally funded research center · nuclear fuel reprocessing

At a federally funded research center: process simulation of nuclear fuel reprocessing — a working digital twin of an industrial process, built years before "digital twin" was a category.

Demonstrations

Proof you can audit.

Client work stays confidential, so demonstrations never ask for trust. When we show what a method can do, it runs on open data or a fully disclosed synthetic system — with the complete pipeline, walk-forward splits, and the naive baselines in the same table. If a result cannot be rerun, we do not show it.

Lineage

Where the craft comes from.

The practice behind the firm spans a doctorate in system dynamics — the discipline of modeling feedback in real systems — and fifteen-plus years of building data pipelines, forecasting models, and simulations in production.