Skip to content

Customer service & order desk · B2B and fleet customers · ERP rollout and support

I ran an office for five years.I know that process from both sides — desk and system.

My name is Patryk Piecyk. I spent 7.5 years at a German company — five of them running its office: orders, invoices, complaints, ERP, B2B customers. Since June 2026 I build tools for that same process. I am looking for a permanent role where both count: knowing the process from the inside and being able to see a tool through to the end.

at a German company
7.5 yearsat a German company
a panel and an iOS app, from scratch
2 appsa panel and an iOS app, from scratch
a day in season — the process under load
40–45 carsa day in season — the process under load
Get in touchSee what exists
Warsaw · available immediately · remote and hybrid · on site in Warsaw only↓ Download CV (PDF)kontakt@patrykpiecyk.pl
About me

Let's start with what I don't have.

If you're looking for a computer science degree and five years at a software house — that's not me. I'd rather say it in the first paragraph than in the third interview. On the right is what I actually have.

⚠

What I don't have

  • ✕A computer science education — general secondary school, matriculation 2017, no university
  • ✕A German language certificate — I learned it at work, not on a course
  • ✕Experience inside a developer team — no code review, no shared repository, no sprints
◇

What I do have

  • ✓7.5 years at a German company (2017–2025), five of them running its office
  • ✓Orders, invoices, complaints — PlusFakt Enterprise ERP and the infra-struktur CRM (NetzKultur), documents into DATEV
  • ✓German fluently, spoken and written; English at a working level
  • ✓Peak season of 40–45 cars a day and a 900-set warehouse — I know how a process behaves under load
Skills

Two things I genuinely know.

Not a buzzword stack. What I did every day for years, and what comes out of that knowledge today.

⌘

The office process: from order to payment

For five years I owned what is now being automated: taking the order, issuing the invoice, handling the complaint, chasing payment, dealing with B2B customers, working in the ERP (PlusFakt Enterprise) and the CRM (infra-struktur by NetzKultur), and passing documents to accounting. I know this process from the position of the person accountable for it — not from a ticket description.

  • PlusFakt ERP
  • infra-struktur CRM
  • DATEV
  • Invoicing
  • Complaints
  • B2B support
  • German
⚡

Building tools in tandem with AI

The design, the decisions, the tests and the responsibility for the outcome are mine; the code is written in tandem with an AI assistant. That is how the business panel with KSeF integration and the iPhone/iPad app came to exist. I state it openly, because it's a way of working — not something to hide in a CV.

  • Next.js
  • TypeScript
  • Postgres
  • SwiftUI
  • KSeF
  • n8n
  • Ollama
What exists

One process, proven four times.

Everything below circles the same thing: the path from order to payment and whatever runs it. Once I ran that path inside somebody else's company for five years, once a system came out of it, once I carried it into the field, and once I did the same for a former employer's process.

Business panel with KSeF integration

Leads, clients, quotes, contracts, invoices, expenses, projects, mail and calendar in one place. Over 240 API routes, custom auth with a second factor, self-applying migrations, a health screen with an error log.

KSeF: “Accepted” status in the Ministry of Finance test environment

iPhone and iPad app

Native SwiftUI, not a hybrid shell: invoices, quotes, calendar, mail, a project timer, notifications and Live Activity. The iPad gets its own two-column layout with Apple Pencil support.

Started from an empty Xcode project

Processes that run every night unattended

The panel's back end doesn't stop at the screen: some jobs start themselves overnight, on my own hardware. What matters more than what they compute is what happened on 1 July — they stalled silently, because one path had no time limit. I found the cause and fixed it so it can't come back somewhere else.

Outage diagnosed and closed — write-up in Notes

Two systems and a website for my former employer

Built for the specific process of a German tyre service I know from the inside: a call register with local speech recognition and conversation summaries, a technical wheel-specification database (PL/DE, works offline) and a company website. Documentation and staff instructions included.

Finished and tested, not yet rolled out — the decision sits with the company

Screenshots

This is what it looks like inside.

Before you click: no, the world does not need another CRM, and I am not trying to sell this one. There is a single reason it is here — to show that I can carry a process from order to payment through a working system: design it, wire it into a government API, keep it alive after an outage, and document it so somebody else could take it over. The screens are evidence, not an offer.

Dashboard: what needs doing today

Overdue receivables, leads with no reply, messages to answer, and suggestions the panel derives from SQL rules — no AI model involved. The decision is always a human one; the panel only asks.

Every screenshot comes from a local test database and from the simulator pointed at the same server — hence “[sample]” and “Jan Przykładowy”. There are no real clients in the panel, because I don’t run a business; what it holds is my own data, and that doesn’t go into screenshots either.

Patryk Piecyk
Who you're talking to

One person. No team, no agency.

Everything on this site I run myself: the design, the copy, the decisions, the testing, the deployment and the upkeep. The code was written in a pair with an AI assistant — I say so plainly, because I would rather tell you than be caught at it. There is no team and no account manager; that isn't a marketing angle, it's simply the fact. I'm currently looking for employment where this combination is useful: knowing the office process from the inside, and being able to get a tool built that runs it.

“I don't have a computer science degree. I have a working system, and I can show you how every part of it was made.”

Patryk Piecyk— Warsaw · open to employment

When a system stalls

A real incident. A real queue. Zero panic.

Every system stalls eventually — the question is what happens next. On 1 July mine did, and it did so silently, with no alert. Below: three numbers, the cause and the fix. On a client rollout this same skill decides whether an outage costs an hour or a day.

Self-healing events

33

Observation window

17 days

Longest wait

78 min

queue released automatically

Root cause

The proxy process hung internally after a run of long requests: it kept listening but stopped processing new ones. The culprit was a dead, orphaned network connection and the absence of any timeout on the critical path.

The fix

Not a restart, but a change in the code: the queue is now guaranteed to be released on EVERY path (success, error, timeout), every step has a hard time limit, and each self-recovery raises an email alert — so silence never means "all good".

Read what exactly happened →
Questions you'd ask anyway

Answered upfront, so you don't have to ask.

The code was written in tandem with an AI assistant. Mine are the design, the decisions, the tests, the fixes after outages, and the responsibility for the thing working. I won't pretend I typed it from memory — but that doesn't mean I don't know what happens inside. Ask me about any part of it.

If this is the profile you need — write.

I reply the same day. Concrete is best: what it's about and which process needs sorting. In Polish, German or English.

↓ CV
☏ PL
Get in touch