PyData Trójmiasto #45

PyData Trójmiasto is back after vacation time! 45th edition is coming up!

Where: Łużycka 8B, Gdynia, Tensor Z building.
When: 30th of September at 6PM

Registration: please register with your full first and last name. Don't forget to bring your ID when entering the event venue.
RSVP available until 6pm day before. In any urgent communication please let us know at kontakt@pydata-trojmiasto.pl

Parking: Free parking located along the street next to the building.

Agenda:

18:00 - 18:05 - Event boarding
18:05 - 18:10 - A few words about PyData
18:10 - 18:55 - From Changelog to Blog - AI Pipeline for Release Communication by Karol Błaszczak
19:00 - 19:45 - Yet Another Year of Agents by Mateusz Hordyński
19:45 - Networking

From Changelog to Blog - AI Pipeline for Release Communication by Karol Błaszczak

Release communication means so many different things to different people. From changelogs and release notes, through in-app guidance and product tours, to release newsletters and blog posts. Who has time for all that, right? AI does. If you use it well. Otherwise, you will likely waste both time and tokens. Let’s have a look at potential benefits of using LLMs to generate release communication channels, and the ways to ensure alignment between documents, with no hallucinations added. We will discuss a pipeline based on a single agent with multiple input sets.

Karol Błaszczak, currently, works on Open Edge Platform documentation and focuses on propagating technical information without the marketing fluff.
Having worked as a teacher, a translator, and a content creator, he prioritizes the language and message clarity, trying to strike the right balance between detail and readability.
After eight years spent in AI-centric Intel projects, he continues to look for the correct path to make LLMs support good content for people, instead of flooding the market with AI slop nobody needs.

Yet Another Year of Agents

A field report from a year of building agentic systems with enterprise teams: what actually survived production and what got quietly deleted at 2am.

Where multi-agent setups are worth the complexity—and where they’re just org charts disguised as architecture. Why evaluation is still the bottleneck nobody wants to talk about. How much of “agent engineering” is really context wrangling and permissions plumbing. And the failure modes that keep showing up, regardless of which model you’re paying for.

Opinionated, occasionally wrong, and based on real deployments rather than demos. Useful if you’re building agents—or about to be told you are.

About Mateusz Hordyński:

Mateusz is a Senior ML Tech Lead at deepsense.ai, where he designs #generativeAI applications, builds large-scale data platforms, and leads projects across data engineering, #MLOps, #RAG systems, and #agentic applications. He’s also the lead maintainer of the open-source project db-ally and one of the people driving technical knowledge-sharing initiatives inside the company.