Should you build it? I'll find out, in phases.
I'm Sergio, an optics physicist who builds AI systems.
- Have a lab prototype? I test what it does and size its market.
- Have an AI idea? I build it: an assistant that answers from your own documents, or an agent that runs tools safely.
In Barcelona. Remote or on site, in English or Spanish.
Stop at any gate. You keep everything done so far.
- First
We talk
About what you have.
You get a written brief -
Phase 1
What does it do?
I take the technology apart on paper: what it measures, how well, and against what.
You get a technical assessment -
Phase 2
Can it sell?
Who buys it, who it competes with, and how big the market is.
You get a market analysis -
Phase 3
How do we build it?
A plan with team, costs and financing. For an AI idea, we start here and I build it.
You get a plan, or the working assistant - At every gate
You decide
with the evidence in writing.
Recent work
- Independent technical consultant, client confidentialIs this prototype worth a product?A new technology, a hard question: is there a product in it? I assessed what it does, sized the market and laid out the next steps as decision gates.
- Optical metrology companyAn assistant for the company's own knowledgeAn internal LLM assistant and a document database, so technical answers come from the company's own product and scientific documents.
- Own project, open sourceA documentation assistant, brief to handoverRun like a client project: requirements, an evaluation set frozen before the first run, a runbook. It got 26 of 31 unseen questions right and declined all 8 off-topic ones. See the repo
- Own project, live replayAn agent runs an optics experiment, safelyA local model drives a simulated bench through tools with a power cap and interlocks. It found the hidden answer in 28 steps. Replay the run