Build
what’s next.
AI & Development Services
Trusted by founders & teams
building what’s next.
Hi, I’m Nicolas.
I build AI systems that turn ambitious ideas into real products.
From intelligent agents to production-ready applications, I help ambitious teams move from possibility to something people can actually use.
What I build
The six problems I get hired for.
AI agents
Software that runs a whole task on its own: looks things up, makes the call, updates your systems.
02Voice AI
An AI that answers the phone, handles the routine calls, and passes the rest to a person.
03Answers from your own data
A chatbot your team can ask instead of digging through the wiki. Built on RAG, so every answer comes with its source.
04The whole product
The app around the AI: accounts, permissions, billing, and the screens people actually use.
05Lower AI costs
Most AI bills are an expensive model doing cheap work. I find where that happens and fix it, usually without users noticing.
06Harness engineering
The rules and checks your team's coding agents work inside, so what they write comes back reviewable at the speed it arrives.
How this works
Five ways in. The less certain you are, the smaller you start.
Every engagement is scoped to a single flow, and one person does all of it: the model work, the product around it, the infrastructure it runs on, and the compliance it has to pass. That is why these take days and weeks instead of quarters. You explain your problem once, to the person who builds it.
A second opinion
90 minutesBring me an architecture, a vendor quote, a stalled build, or an AI bill that got out of hand. You get a written verdict, not a proposal: what holds, what does not, and what I would do instead. The verdict can be: do not build this.
Scoping
1–2 hours · fixedOne or two hours on a call, working out what to build against your real data, your real constraints, and the rules on where that data may live. You leave with a spec precise enough for anyone to build from: me, your own team, or the agency you are comparing me against. Every step after this one is fast because this happened first.
Prototype
1 week · fixedA working prototype on your data, in front of your stakeholders, one week after scoping. It either unlocks the budget for the real build or kills the idea before it costs you a year.
Build it properly
2–8 weeksThe full system in production, used by real users. Evals before launch, tracing you can read, a cost per request you can defend to finance, and a handover that leaves your team able to change it without me. Steps 01 to 03 are cheaper ways of deciding whether to do this one.
Stay on
MonthlyI stay responsible for your AI systems: building new capability, reviewing what your team ships, and keeping cost and compliance in check as the product grows. You get a senior AI engineer without having to hire one.
If none of these fit
Most work lands on one of these five. If yours does not, bring it to the call anyway and I will tell you what it needs.
The next step
Put thirty minutes on the calendar.
A free half hour on what you are building and whether I can help. Every way in above starts with this conversation. Pick a time; the invite lands in both calendars.
Not ready to pick a time? Ask the assistant something awkward, or write to [email protected], and I answer that one myself.
— Nicolas