PLEN

Jakub Chodnik

Process automation and AI implementation

I design the logic and architecture of systems and use AI to build them.

Currently
kreatywnylas.pl
Location
Katowice / Kraków, Poland

About

Jakub Chodnik

Since September 2025 I have been working at kreatywnylas.pl, an e-commerce company offering personalised products and printing. Alongside my job, I am studying Computer Science and Business.

I started out writing FiveM scripts (2020–2023) and Minecraft plugins (2023–2025). Today, what I enjoy most is the moment a manual process simply disappears.

I want to design increasingly autonomous agent systems: ones that run an entire process on their own and bring in a person only where one is truly needed. The first step is an AI agent in the message centre that already writes replies to customers. For now, I review each one before it is sent.

University
Andrzej Frycz Modrzewski Kraków University
Studies
Part-time,
Oct 2025 – Mar 2029 (expected)
Languages
Polish, English (advanced)
AI & models
Claude Code Anthropic API ChatGPT, Codex and OpenAI API MCP Ollama Qwen
Automation & web
n8n WordPress Vercel
Data & deployment
Supabase PostgreSQL Docker GitHub
Skills
AI agent architecture and development, local model deployment, print production automation

Experience

kreatywnylas.pl

September 2025 – present

AI and IT: developing solutions based on artificial intelligence · full-time

  • I develop the message centre where the team handles customer messages from Gmail, Allegro, Instagram and Facebook. I designed its new interface and an AI agent that drafts replies to customers using a local Qwen model.
  • I built a panel for the production team that generates print-ready files from order lists, and I keep adding new products to it.
  • I use n8n to connect the company's systems with other services, including Instagram, Facebook and Sellasist, and to automate repetitive customer service tasks. See Projects for details.

Selected projects

Conversation list from Gmail, Allegro, Instagram and Facebook 1 / 6

Screenshots use sample data.

Message centre expansion and AI agent

Before
The core of the customer service system (receiving messages and replying from the panel) was built before I joined the project.
What I did
Since July 2026 I have been developing the system. I designed and built a new interface for the panel and an AI agent running on a local Qwen model. The agent writes a reply to every customer message using order data and the shop's product pages, asks for missing personalisation details and flags risky cases, such as complaints, for review. I also added Allegro complaint handling, Instagram, Facebook and Sellasist integrations, tags, an autoresponder and alerts for unanswered conversations.
Now
I review the agent's replies before they reach the customer and will let it send them on its own once their quality is good enough. Corrected replies are fed back to the agent as examples. I keep adding new features.

Deployed, under active developmentNode.js, PostgreSQL, n8n, Docker, Qwen

Production: order data and PDF generation 1 / 4

Screenshots use sample data.

Production and team communication panel

Before
Print files were prepared by hand, which was slow and error-prone.
What I did
I built an internal web app for the production team, hosted on the company network. It generates print-ready files for Roland VersaWorks straight from order lists, for products including inserts for CD boxes, calendars, cups and mugs. It also includes a reprint queue for complaints and an internal team chat. I keep adding new products so that new items go straight into production, with no print files prepared by hand.
Result
Mistakes have been cut to a minimum, and preparing production takes far less time.

Deployed, under active developmentPython, Streamlit, PyMuPDF, ReportLab