There is no money to pay anyone yet, so nothing below is a job posting. What is real is that the right person changes what this company can attempt, and those conversations are worth having before there is a budget rather than after.
Co-founder / CTO
Open conversationThe person who owns whether this actually works.
Take the private training benchmark from a research script to something a research group runs on its own hardware, behind its own firewall, without us present. Own the honest answer to how much pooling really buys. Own the security posture — the whole claim is that the data does not leave, and that claim will be tested by people paid to test it.
What we would look for
- —Built and shipped distributed ML systems that ran somewhere other than your laptop
- —Depth in federated learning, secure aggregation, applied cryptography or privacy-preserving ML
- —Comfortable as the most senior engineer in the room with nobody to escalate to
Quantum background is not required. The quantum layer is years out.
Chief Scientist
Open conversationThe person who owns the scientific question the company rests on.
Define the benchmark — what task, what datasets, what baseline, what result would falsify the thesis — and write it down before running it. Establish the honest delta between a federated model and a single-silo model on real materials data. If that number is small, the company needs to know in month three, not year three.
What we would look for
- —Track record in computational materials science, chemistry or scientific ML, with publications we can read
- —Still writes code and runs experiments
- —Rigour about negative results — the most valuable thing you could do in year one is establish that the effect is real, or that it is not
If a full-time move is not possible, an advisor arrangement is a real alternative and we would take it gladly.
Chief Commercial Officer
Later — after the benchmarkThe hardest problem here is the first three participants, and it is commercial, not technical.
The technology can be demonstrated on public data. What cannot be demonstrated in advance is that a research director will put proprietary experimental data into a training run operated by a startup. Build the buyer map, run the conversations that validate or kill the thesis, land the first two participants.
What we would look for
- —Sold complex, high-consideration deals into R&D or technical buyers
- —Opened a market that did not have a category yet, and can describe how
- —An existing network in materials, chemicals, batteries, semiconductors or pharma R&D — close to decisive, because the founder's is honestly zero
Overstating a guarantee to close a deal would end this company. That is not aggression here, it is a misrepresentation in a signed contract.
Chief Operating Officer
Not yet — roughly ten people outListed for honesty, not because we are hiring it.
A pre-revenue research company with one person does not need a COO, and hiring one early is a recognisable failure mode. When it exists, the job is consortium operations: onboarding a participant is a six-to-twelve-month process across their legal, security, IT and research teams, and making that repeatable is the difference between a company that scales and three heroic deals.
What we would look for
- —Ran operations at a company selling into large, careful, regulated organisations
- —Personally negotiated complex commercial contracts, not just approved them
- —Consortium or joint-venture experience — rare, and close to decisive
If you join and there is no operational load yet, you will be underemployed and you will leave. Better to have that conversation now.
A no arrives by the end of the week. A slow no is the most expensive thing a small company does to its reputation.
