Why I Formed Parametrica

I spent a decade building the models, and then three years sitting where the decisions get made. The two almost never met. Parametrica exists to close that distance. This is the honest version of why, written while the company is still small enough for the reasoning to matter more than the record.

First I built the models

A PhD at Imperial College London, then six years at the National Renewable Energy Laboratory in Colorado. High-fidelity simulation of the atmosphere, the ocean and the structures we put in both. I directed a $3.9M US Department of Energy programme on hurricane impacts on offshore wind, and our turbulence-resolving simulation of Hurricane Laura reproduced the storm's extreme winds and eyewall turbulence directly from the physics. [1]

The models work. That is not the interesting part. The interesting part is what happened when I left.

Then I sat on the other side

I went into industry and led renewable-energy and offshore risk model product lines at Moody's, and then offshore wind product at a specialist risk analytics firm. That is a different room. In that room nobody asks which solver you used. They ask where the number came from, what it assumes, and what happens to the answer if the assumption is wrong.

Often the honest answer is that the number came from a fleet average, that it assumes this site behaves like the fleet, and that nobody has checked. Everyone involved is competent and nobody is being careless. The number is simply the best one available on the day it was needed.

The two never met

Renewable energy projects rarely get into trouble because someone got the physics wrong. They get into trouble because a number could not be defended when it mattered. A developer needs a number to close financing. An underwriter needs one to price the risk. A lender needs one to release capital. Underneath each of them sit assumptions about the governing loads or the site conditions, made early and rarely stress-tested again.

The physics that would test those assumptions exists and it is good. It arrives too late, or not at all, because it is slow, expensive, and lives in institutions whose job is not to meet a bid deadline. Nobody designed this gap. It is just what happens when the people who can resolve a question and the people who have to answer it work in different buildings on different clocks.

So this is an R&D company

Parametrica is a research and development company. Much of what it sells looks like due diligence, because that is what this market actually needs to buy: a number checked, or challenged, by someone independent. The difference is the standard it is held to. What makes a number defensible is research discipline rather than seniority, traced back to the physics that produced it, tested across the uncertainty around it, and documented well enough that a stranger can re-run it and disagree with you properly. The role I want is the one a project calls when a number has to hold: technical due diligence, done from the physics up.

Underneath sits one modelling chain. It starts in the atmosphere and ocean and nests down into turbulence-resolving large-eddy simulation, because the eddies that actually load a machine have to be resolved rather than approximated away. It passes through the structural and aeroelastic response of energy infrastructure onshore and offshore, uses machine-learning surrogates where speed is the binding constraint, and carries uncertainty the whole way, so that a band we publish is a band we have measured.

That one chain answers more questions than I expected when I started. What the wind does over a site, and how much of it the turbines take from each other. What a storm does to a specific machine. How a component ages. And, underneath all of them, whether the data any of it rests on is good enough for the decision being made, which is a question almost nobody asks in the units of the decision itself. Getting the chain into a project early is as much a workflow problem as a physics one, which is why we build it around TurbineX, a solver-agnostic platform that folds engineering-tier and high-fidelity models into the same design loop.

The same work gets sold twice. As work packages inside European research consortia, which fund the depth and force the results into the open literature. And as fixed-scope studies for industry, which keep the work pointed at real deadlines and real sceptics. Neither half is a side project. A research company with no commercial deadline drifts, and a consultancy with no research loses the thing that made its numbers worth buying.

What I am trying to prove, and how you will know if I am wrong

The bet is that high-fidelity physics can be delivered on a commercial deadline, at a price a project will actually pay, and that when it can be, people prefer it to a fleet average.

That is testable, so here is the test. If in two years this work is still bought only by people who already believed in the physics, and never by someone who simply had a deadline and a sceptic to satisfy, then the bet was wrong and the honest conclusion is that the industry is content with inherited numbers. I do not think that is true. But I would rather state the test now than claim a track record I have not built yet.

An invitation

If you are holding a number you would not want to defend in front of a lender's independent engineer, that is the conversation. Developers, consultants, certifiers, underwriters, and research groups assembling a consortium: I would like to hear what the number is and what it is resting on.

References

[1] Sanchez-Gomez, M., Deskos, G., Lundquist, J. K. (2025). Turbulence-resolving simulations of Hurricane Laura (2020): Insights into extreme winds and eyewall turbulence. Quarterly Journal of the Royal Meteorological Society, 151(773), e70003. 10.1002/qj.70003