Open to B2B contracts · remote from Wrocław
An AI product engineer who owns the feature, not just the ticket.
Senior product engineering on a B2B contract for AI-native teams in Europe, the UK and the US. Nine years shipping for companies in Germany, the Netherlands, the UK and the US, and two of my own AI products live in production, built solo with Claude Code and Codex in the loop. I take a feature from spec to production and own what happens after it.
What I bring to a team
AI features in production
LLM features wired into a real product, not a demo: structured output and tool use, RAG, matching and ranking, eval sets that catch regressions before users do, and the cost caps, rate limits and kill-switch that keep an AI endpoint from becoming an open wallet.
- Structured output · tool use · RAG
- Golden-set evals
- Cost caps + kill-switch
Product engineering, end to end
Next.js and TypeScript on the front, NestJS and PostgreSQL behind it, deployed and instrumented. I take a feature from the spec to production and read the numbers after release, because a shipped feature that nobody uses is not done.
- Next.js 16 · React 19 · TypeScript
- NestJS · PostgreSQL
- PostHog · error tracking · SEO/GEO
Agent-native delivery
Claude Code and Codex CLI with the guardrails that make their output mergeable: written specs, engineering standards the agents are held to, deny rules and hooks, a second-agent review and my own review of every pull request. I can set that workflow up for your team, not only use it.
- Claude Code · Codex CLI · Kiro CLI
- Specs, hooks, deny rules
- Human review on every PR
How we would work
B2B contract, no entity needed
I invoice as an independent contractor from Poland, so you do not need a Polish entity. A long-term engagement inside your team, or a well-scoped build with a clear finish line.
Your time zone, in writing
Remote from Wrocław (CET): full overlap with European teams and a shared half-day with the US East Coast. Async by default, with specs, pull requests and decisions written down.
Fast, because of how I work
AI agents handle the end-to-end volume while I own the judgment: architecture, product calls and verification. That is what turns weeks into days without turning quality into slop.
Shipped and measured
Analytics, error tracking and tests on the paths that break money or data come with the feature, not after it. You can see that it works, and so can the next engineer.
From the team I shipped with
Three recommendations from my team at Maandag in the Netherlands: the product owner I reported to, our business analyst and a fellow developer. Quoted word for word from LinkedIn.
Piotr worked as Front-end developer in my team at Maandag. I enjoyed working with him and he was an important part of the team, helping to realize our digital ambitions. He works well with other team members, takes responsibility and has a focus on quality. I would recommend to hire him for your team.Joost Verbiesen
Product Owner · my manager, Maandag
LinkedIn · Oct 2023 ↗Having Piotr in your team, is making sure that work is delivered of top notch quality. He doesn't shy away from questions or confrontations, which makes it very pleasant to work with him, as he is able to challenge others. Piotr is very eager to learn and improve himself and cares about your product. I would definitely like to work together with Piotr again.Sherilyn Pang
Business Analyst · Maandag
LinkedIn · Oct 2023 ↗I had the pleasant opportunity to work alongside Piotr at Maandag. He has a professional work ethic, a commitment towards his work and craft, is cohesive within a team, is always willing to lend a helping hand, and always a pleasure and inspiration to have a brainstorming session, consultation and discussion. I would highly recommend Piotr!Craig Spindler
Developer · Maandag
LinkedIn · Oct 2023 ↗