Abstract: Join Jan Soubusta for a tour from SQL-only analysis through today’s drag-and-drop dashboards to reliable AI assistants.
See how GoodData’s AI stack (Qdrant, small LMs, rerankers), an LLM-as-judge layer, and other tools make analytics invisible.
Finally, learn what we are going to build to support reliable autonomous AI agents (knowledge graphs, MCP, A2A).

About: Keboolian, father, lecturer, mentor, event organizer, community enthusiast,  author. 2 decades pushing data from all angles, couple years in product management and most recently responsible for Product Marketing Engineering in Keboola.

Abstract: We all know the guilty pleasure of “just hacking it together” to see if things work. In data engineering, that usually sounds reckless — pipelines break, configs get messy, and tomorrow you regret it, like a letf em dash in a session description.
This session argues the opposite: that a bit of vibe coding is not only possible, but can make your daily work faster and clearer. We’ll show how building and debugging pipelines can be done in minutes, without drowning in configs or waiting for heavyweight stacks like Azure, Google, or Databricks to catch up.
It feels dangerous, it looks magical — but in the end, it’s a safe, practical way to run a simpler data platform.
Session plan:
1) Intro, MCP Server is not just an API wrapper, what’s possible. (~20 min)
2) Live demo (~20 min)
3) Let’s try it yourself. (~40 min)