A model that turns a big, ambiguous infrastructure question into the one number you can actually decide on.
$87M. The upfront cost gap between two ways to fuel the same 25 MW site, found before a dollar was committed, and a premium the model showed would never pay back.
The decision: how to power a roughly 25 MW behind-the-meter data-center build in northern Indiana, co-located with utility-scale solar. One live question split the architecture in two. Bring natural gas in by pipeline, or build on-site gasification and run the same generators on syngas. Same generation, same site, same round-the-clock load. The two paths sit tens of millions of dollars apart, and the choice gets poured in concrete on day one.
The expensive mistakes in infrastructure happen before anything is built. The debate in the room was about fuel: which feedstock would be cheaper to run over the life of the plant. That is the assumption everyone argues about, and it is the wrong one to anchor on. The real risk was sitting somewhere nobody was pressure-testing.
Deep research grounding first, then a model that sizes the whole power system for the site and tests the decision against the assumptions that actually move it. Instead of a single point estimate, it re-runs the build across the full realistic range of every input that matters, fuel prices high and low, on both sides at once. It also tells you what it will not tell you: it reports ranges and shows where the outcome is most sensitive, not a single right answer, because the decision does not have one until the site inputs are locked.
The split came down to one line item. On-site gasification carried about $88M of upfront infrastructure cost. The pipeline carried roughly half a million. Everything else about the two builds was nearly identical. Then the model tested whether that premium could ever pay back through cheaper or greener fuel, and re-ran the decision across the full realistic fuel-price range, on both sides at once. The pipeline won in every case. The fuel-price debate, the thing everyone was arguing about, never once changed the answer. The decision was never about fuel cost. It was about the capital that nobody was stress-testing.
What this shows about working with me
This is the same first-principles move I bring to a stalled build. Find the variable that actually decides the outcome, separate it from the one everyone is arguing about, and make the call defensible before the money moves. Here the domain is power. The muscle is diagnosis, and it is the same one that rescues a delivery that has lost the plot.
