AI engineering meetup
Deploy AI
#3 · Safety and Efficiency in AI Engineering
Deploy AI #3
Tuesday, 20 October 2026 · 18:00
Upside office, Krupnicza 5/6, Kraków
About the series
A space in Kraków for engineers, where LLM demos meet production.
Deploy AI is a meetup for engineers who put LLMs and agents into production and then have to live with the consequences, the struggles and the wins. We skip the demos and talk about the problems that come later, and about the people dealing with them.
Every talk takes about 25 minutes, followed by questions. The content goes deep technically, and sometimes it is controversial enough to keep everyone talking long after the Q&A.
If you build software with AI every day, or you are about to, come by. If "prompt engineering" already bores you, even better.
Practitioners on stage
Speakers who have built it, broken it and fixed it.
Room for questions
Every talk has Q&A, and disagreement is welcome.
Networking
A room of 20–35 people working on the same problems as you.
See what a Deploy AI evening looks like
Latest edition · Deploy AI #3
Safety and Efficiency in AI Engineering
Two talks on working with LLMs efficiently and safely: naming things precisely to get better answers, and giving coding agents autonomy without handing over the keys.
- EN
Learn the Taxonomy First: The Key to Efficient LLM Use
The key to using LLMs efficiently is learning the taxonomy of a field first. Precise names lead to precise answers: faster, sharper, and with fewer corrections.
Wiktor Daniec
Senior Software Engineer, Upside
- EN
Sandboxing Claude Code: Running AI Agents Safely
AI coding agents are most useful when they run autonomously, but every skipped permission prompt hands more control to a model that can make mistakes or be manipulated.
Krzysztof Pokorski
Software Engineer, Upside
- Date
- Tuesday, 20 October 2026
- Time
- Doors 17:45 · Talks 18:00
- Venue
- Upside office, Krupnicza 5/6, Kraków
- Entry
- Free · registration required
Call for speakers
Built something worth talking about?
We are always looking for speakers. You do not need a polished deck or a list of past talks. You need something you have actually built, broken or figured out, and the willingness to talk about it honestly with a room of AI engineers.
Not sure your topic is big enough? A narrow topic is fine. One specific problem explained well beats a broad overview.
$ ./submit-talk
Archive
Past Deploy AI talks
Every talk so far came from someone working on the problem, not just reading about it.
- #2 · 8 Sep 2026
From Neurons to Transformers: How We Taught Machines to Read
Paweł Wiercioch
Software Engineer, Upside
- #2 · 8 Sep 2026
Why AI Made Neovim Worth Learning
Wiktor Daniec
Senior Software Engineer, Upside
- #1 · 7 Jul 2026
AI Hacking: Doing it the Wrong Way is the Best Way
Maciej Majewski
Lecturer, AGH
- #1 · 7 Jul 2026
16-Hour Tasks: How to Make AI Handle Long-Horizon Work
Wiktor Daniec
Senior Software Engineer, Upside
