SLO AI Lightning Talks #2
Fast-paced talks on voice agents, spec-driven development, Slovene LLMs, empathetic customer support, visual similarity search, and data loops.
SLO AI Lightning Talks #2
Date: Thursday, December 11, 2025 at 18:00
Location: Povio, Ljubljana (BTC)
Registration: Early access for SLO AI members, general sign-ups open next week
About This Event
SLO AI Lightning Talks #2 is just around the corner, bringing a fast-paced lineup of talks covering voice enabled agents, spec driven development, open source Slovene LLMs, empathetic customer support systems, visual similarity search, and data loops. The evening will be hosted at Povio, with support from Codeplain and Celtra, whose backing makes gatherings like this possible.
General sign-ups open next week, but SLO AI individual members get early access before registration goes public. If you're already a member, you'll be able to secure your spot ahead of everyone else.
Want Priority Access to Events?
Season 2025/26 Membership (€19/year) - Support Slovenia's AI community and get exclusive benefits:
- Priority/early registration for all events (events remain free!)
- Early updates on new initiatives and job opportunities
- Direct support for the local AI ecosystem
Valid from October 2025 to October 2026.
Get Your Season Membership Here
Meet the Speakers
LiveKit: Make Agents Hear & Speak

Go beyond text-based AI. In this talk, you'll learn how to use LiveKit to give your agents a voice, choose the right LLM and voice stack, and build reliable, real-time AI experiences.
by Andraž Pajtler
Mobile expert with 10+ years of experience. Built apps used by millions across sports, entertainment, logistics, and health. Currently at Reveri, developing an AI agent for personalised hypnosis.
From Vibes to Specs

Lessons learned from building the same tool twice - once through vibe coding, and once through spec-driven development with codeplain. This lightning talk compares the two approaches and shares practical takeaways from doing both.
by Tjaž Eržen
Founding engineer at codeplain and co-host of the Ogrodje podcast. Loves working where AI, engineering, and product development meet.
GaMS – Large Language Model for Slovene

Meet GaMS, the state-of-the-art open-source large language model for Slovene. Domen will outline the main challenges and benefits of training such a model, and introduce the Slovene LLM Arena for open benchmarking and evaluation. With GaMS, Slovenia is at the forefront with one of the EU's most advanced open-source LLMs.
by Domen Vreš
Researcher and PhD student at the University of Ljubljana, Faculty of Computer and Information Science. Lead developer of open-source large language models for Slovene.
I want to talk to a manager - not an AI!

Customer support bots often fail because they miss the emotional context of interactions - how do we turn a frustrated customer into a satisfied one? In this talk, we'll take a look at detecting emotions & generating genuinely empathetic responses using DSPy architecture inside a Sveltekit web app.
by Blendor Sefaj
Co-founder of Clover Labs, a digital product studio. Previous experience includes companies as Creativio, Bitstamp, Enya Labs; he co-organizes tech events Web Hours Maribor. Interested in the intersection of technology, entrepreneurship & humanities, focusing on applicable use of AI in these areas.
Building a Visual Similarity Search from Scratch

Robert will present a hands-on demo on building a visual similarity search system using a modern stack. We'll explore whether an image can be used to identify a person or object across other images, and discuss the technology that makes it possible.
by Robert Sedovšek
Developer with a PhD in UX, exploring ML systems for facial recognition and image similarity. Former manager at Turtl & developer at Celtra.
Data Loops and Data Sovereignty

How to best integrate LLM with data while ensuring the long-term availability of high-quality datasets? Aleks will outline the principles and concepts of data sovereignty, in which computation moves to the data rather than the other way around. The current industry cycle of boosting computational capacity will be followed by a new phase of boosting data - something we can advance together.
by Aleks Jakulin
Winner of the EurAI prize for the best PhD dissertation in AI in Europe (a few years ago) - now working on a new startup, data.flowers.
Registration
Open access on Eventbrite opens next week, with early access available to those who joined through the membership link above.


