First of all, thanks a lot @Gergely Daróczi for inviting me to join the Spare Cores team at the conference (thx for the talk SELECT instance FROM cloud WHERE workload = ? ORDER BY cost_efficiency too, the topic was very close to me :) )! It was an exciting opportunity to attend such a great event. I gathered a lot of ideas, cloned a lot of repos, and as an extra I learned couple of things about myself, which I don’t think is a common experience from an IT conference :).
After a career change I’ve been a developer for about 3 years. PyData was my second conference in my life, so there were a lot of common libraries in the tech stacks whose purpose I didn’t really know (for example Jupyter; I’d never used notebook frameworks before, just mentioning the most trivial example). So many thanks to all the presenters, I could learn something from every talk and workshop I attended.
As an environmental engineer (though I’ve never worked as one) and a map fan, I really liked Mapping the local heat transition: from large-scale geospatial data to real-world impact by @Sofia Pinto and Simran Dave, and The Clean Energy Graveyard: Using Python & Gemini to Map the UK's Cancelled Renewables by @Damian Bemben. Processing geospatial data and messy documents are two areas where AI-related tools have proven genuinely useful. These talks gave me a push toward my long-delayed map-related pet project as well, maybe I’ll need to use GeoPandas too.
MCP, or not MCP by @Neal Richardson was very helpful for me. The usability of MCP servers is a very hot topic for me now (maybe for everyone). In some cases MCP may not be the right tool, but in many others it’s very useful. MCP is not dead :)
My favourite talk was about AI burnout: The Human-in-the-Loop is Tired by @Laura Summers. After so many talks about AI agents, I was getting exhausted from the topic, and this talk brought me peace of mind on it, thx a lot! I think I’m a little bit old-fashioned about agent usage (just a chat window in Cursor, I need to understand generated code too, “regulate” AI agents with MCP servers and don't let the agent cruise on my computer with full permissions), and it turns out I’m maybe not in the minority, there is a path between vibe coding and AI Luddite, it was like a psychology session for me!
Finally, a huge thank you to the PyData London organisers and volunteers!