For teams · embedflexible format

I’ll embed in your AI team — a few days a week, or end to end

I bring AI systems from chaos to working software in weeks, not quarters. I take a stalled project or an empty slot in the architecture and turn it into a reliable system — with metrics and GDPR. 15+ years of engineering; AI assembled into a controlled pipeline, not “I use a model”.

I embed in the teamAI from chaos to working product15+ years engineering

verifiedspeed proven on the flagship — six products in ten days, every number opens in code

Working formats

Not a title — a format

You can embed in different ways — from a light start to end-to-end delivery. Brightness encodes depth of involvement. The deep formats aren’t taken cold: we start small and scale up.

0

AI system review

in days

A map of your AI: which models you run, where the risks are, what they cost to operate, and where data-protection law (GDPR) bites. A light start — understand the scope before any commitment.

1

A few days a week on the team

1–2 days/wk

Embedded part of the week: I lead the AI track, hold the architecture decisions and the quality bar, and stay reachable between tasks. As the team grows I’m a multiplier, not a bottleneck.

2

End to end for 2–3 months

2–3 months

I take a stalled project or an empty slot in the architecture and bring it to reliable operation: design the AI system, set the metrics and the method, and train the team to carry it on.

3

Contract-to-hire

to staff

A few days a week and end-to-end work are the main formats, but I’ll discuss a strong full-time offer too. Contract-to-hire is an honest way to test the fit in practice.

These are working formats, not prices. Project ranges live in ways to work together.

Role

I close the gap between business, AI architecture and product

One role, really, approached from two angles — depending on where the team has a gap. It’s not a menu: I take the angle that solves your actual problem.

Tech lead & AI architect

AI that answers from your data; workflows where AI handles specific steps under clear rules; the right model per task; security and reliability under load. For when a pilot needs to grow into an architecture you can trust with real traffic.

Product owner for AI

Business process → architecture → hands-on build → measurable result. For when an AI idea needs to become a working product, not a pretty demo.

What I don’t take is below.

Proof

For each role — a checkable run

Not “I can do AI”, but a concrete outcome with an open number. The flagship opens in code right now; metrics from company work are on my CV — I name the companies and periods and verify on request.

  • AI in production, with metricson CV

    A business loop: request handling 30 min → 2 min, −3h of a dispatcher’s manual work a day

  • Cost & architectureon CV

    The right model per task: AI running up to 5× cheaper than one expensive model for everything

  • Scale & reliability under loadon CV

    A real-time product: 0 → 400,000+ users, response time −90%, cloud migration in 48 hours

  • Idea to producton CV

    A B2B service: idea → $375K early-stage raise (raised, not revenue)

  • End-to-end speedin code

    Flagship: 6 products in 10 days, 145 commits — opens in code

verifiedthe flagship case opens in code; metrics from company work — I name companies and periods on my CV and verify on request

Boundary

Where I’m not the best choice

An honest filter saves both of us time — and builds trust. If your task is in the right column, better to say so up front.

proven

I take

  • AI from chaos to reliable operation.
  • Architecture, the cost of running AI, and reliability under load.
  • Ownership of the whole product — with a measurable result.
  • End-to-end speed — weeks, not quarters.
won’t claim

I don’t take

  • Research on AI models — training from scratch.
  • Project management with no product ownership.
  • A CTO title with no hands-on work.
  • Maintaining legacy code with no product to grow.
One person, two angles — engineering leadership and product ownership. I take the one that closes the gap; if someone else would close it better, I’ll say so.
Next

Three doors into the evidence

Everything above rests on a real run. Three doors — each answers a different question.

Let’s discuss how I’d fit in

We’ll start with a short look at your AI task — no commitment. We’ll pick a format: a review, a few days a week, end to end for a couple of months, or contract-to-hire. A strong full-time offer is on the table too.