Hello!
We invite you to the 33rd meeting of the Tricity Data Community Group!
The meeting will take place at the Hapag-Lloyd office in Gdańsk.
During the meeting, we will learn how to design context for LLMs and automate its refinement, as well as how to process only what has changed regardless of the technology used.
Participation in the event is, as always, free of charge. However, for logistical reasons, we kindly ask you to register via Meetup and arrive before 6:00 PM.
Event details
Date: Wednesday, June 24
Time: 6:00 PM
Location: Hapag-Lloyd Office, Biuro Alchemia IV – Neon, Grunwaldzka 413, Gdańsk, Oliwa – 7th floor.
Agenda
6:00 PM – 6:15 PM Introduction and organizational matters
6:15 PM – 7:15 PM Hubert Jegierski – A prompt is not enough: how to design context for LLMs and automate its improvement
If you’re using LLMs, you may feel that a good prompt works great… until you switch models, the provider deprecates the version you’re using, or further tweaks start breaking what already worked.
In this presentation, you’ll see why “let’s just write a better prompt” is often not enough and where context engineering begins. You’ll learn about prompt technical debt, the impact of context on response quality, and why manually refining instructions quickly stops scaling.
At the end, you’ll see DSPy – a tool that helps reduce dependence on a single model and automatically search for better prompts in a more measurable way.
7:15 PM – 7:25 PM Q&A
7:25 PM – 7:35 PM Break
7:35 PM – 8:35 PM Piotr Tybulewicz – How to process only what has changed – a technology-agnostic pattern
In Fabric or Databricks, we already have built-in mechanisms for incremental processing today, and they are often sufficient. But what if, for some reason, we cannot use them? Or we want an approach that works independently of the technology?
In this session, you will see a universal change-detection pattern that works in any environment — Azure, Fabric, Databricks, dbt, Snowflake, and even on-premises. With it, you can skip data that has already been processed and focus only on what has actually changed.
8:35 PM – 8:45 PM Q&A
8:45 PM – 10:00 PM Pizza & Networking
Our speakers:
Hubert Jegierski
A Machine Learning Engineer with 8+ years of experience in developing and deploying AI/ML-based solutions, combining academic and industry expertise. He has worked on AI/ML applications across a wide range of fields, including natural language processing (NLP), predictive maintenance, anomaly detection, and biomedical data analysis. Currently, he focuses on LLM-based knowledge management, such as fact verification in documents, sales support, and information retrieval.
Piotr Tybulewicz
A seasoned expert discussing data and broadly understood data engineering (especially in Azure).
Submit your topic!
Do you have knowledge you would like to share with others? Would you like to explore an interesting topic, present a case study, or try yourself as a speaker? Get in touch with us – we offer full support!
Email us at trojmiasto@datacommunity.pl.