Team
Three founders, three parts of the problem.
There are three jobs in this company: build the models, know the industry, and sell it. One of us owns each one outright. That is how three things move at once without a meeting about it.
San Diego, California
Chief Executive Officer
Cameron Souza
Modelling, biogeochemistry and aquaculture. Cameron owns the engine and the science. He wrote the domain models behind every product, the estimator that runs on top of them, and the benchmark harness we use to check our own work.
He will tell you the filter is the easy part, textbook maths. The hard part is describing one particular living system accurately enough that the filter has something true to work on. Everything else follows from getting that right.
Co-founder
Zachary Cowden
Biology and industry. Previously at Leidos, working on biomanufacturing. Zachary runs the bioprocess, digestion and life support domains, which are the three where the customer is a regulated programme rather than a farm.
He knows what a programme office will ask for before they do.
Co-founder
Bernard Yong
Markets and operations. Previously at AWS, a lawyer by training, and a former adviser on United Nations policy. Bernard runs commercial structure, channel and the legal interface.
How we work
Four habits we are not willing to trade.
Every one of these has already cost us something. That is the only reason to believe we actually mean them.
- We publish the misses. Every evaluation reports where the engine would have been wrong on that operator's history, alongside where it would have been right. We have thrown out our own headline result before, when the scoring behind it turned out to be improper.
- No number without a source. Every figure on this site resolves to a benchmark, a published paper or a reported operating record, with the run counts behind it.
- Read-only by design. The engine estimates and forecasts. It touches no control loop, which is why an operator can evaluate us without putting anything at risk.
- The operator owns the data. Evaluation files are deleted afterwards, and nothing becomes anyone else's training set without an agreement that says so.
Working with us
Who we want to hear from.
We hire out of conversations rather than job listings. The note we read twice comes from someone who has either written mechanistic models of a living system, or been the person called at three in the morning when one of them went wrong. Those are the two halves of this problem and both are rare.
If that is you, write to Cameron directly and say what you have built or what you have had to fix. That opens more doors here than a resume does.
Send us a log
A free, blind evaluation against your own recorded incidents. No hardware, no install, no connection to a live system.
Read the record first
The validation table on the home page names every benchmark and the result against it. Start there rather than with a deck.
Break our benchmarks
The harness, the preregistrations and the scoring are available on request. We would rather be corrected early.
Just write
cameron@orbitalecology.com reaches a founder, not a queue.
Contact
One email reaches all three of us.
Whether you run a facility, write models, or are working out whether this applies to you, the fastest route is a direct note. We read and answer them ourselves.
- Operators: a free blind evaluation on your own history.
- Researchers: the benchmark harness and preregistrations on request.
- Everyone: no queue, no form, no sales sequence.
Orbital Ecology, San Diego, California. Validated against the reference models these fields use, and backtested on real commercial operating history.