31st Belgium NLP Meetup

It's back to school week, and that means meetups, too, spring back to life. With agents that escape their sandboxes, a blistering pace of model releases, and a couple of landmark IPOs looming, NLP and AI enthusiasts certainly have enough to talk about.

The next Belgium NLP Meetup will take place on Thursday October 8th in the offices of Trendtracker in Ghent. We'll open the doors at 7pm for pizza, have three interesting talks between 7.30pm and 9pm, and after that there's more time for networking and drinks.

Below are the three talks for the evening:

Relex: Relationship Extraction at Scale for Trendtracker's Knowledge Graph
François Remy (Trendtracker)
What can you do when you need to extract relationships at such a scale that LLMs stop being economical? In this talk, François will explain how Trendtracker trained its own multilingual relationship extraction model to power its knowledge graph, which jointly identifies entities and links them by relationship type. He'll discuss why they chose a bi-affine architecture, and how scale, cost, and reproducibility shaped the design. He'll also show how the extracted relationships can be leveraged to help with entity resolution.

Open Source the LLM Engineer, not just the LLM
Sofie Van Landeghem (OxyKodit)
This talk starts rather innocently introducing nanochat: a fully open,
human-readable implementation of a simple LLM with GPT2 capabilities. Sofie then uses nanochat as an example to demonstrate how a bigger LLM can turn into an LLM engineer in an "autoresearch" loop. Finally, she'll show an early prototype of such an open-source LLM engineer, built by a human LLM engineer (Sofie), let loose on a GPU cluster to perform genuinely open AI research.

Beyond CLAUDE.md & .cursorrules: Architecting GitOps and Context Governance for AI Coding Agents
Younes Baghor (SPRAWL.software)
Autonomous coding assistants like Cursor, Claude Code, Codex and GitHub Copilot have dramatically accelerated developer velocity. But as engineering teams adopt these tools across complex codebases, they are hitting a frustrating architectural wall: "Agent Sprawl." In this talk, Younes Baghor, Head of Agentic AI, Systems Architect, and the Founder of SPRAWL.Software, explores how to apply traditional software engineering discipline to AI coding workflows. He’ll do a technical deep-dive into replacing fragmented, ad-hoc prompt files with a unified GitOps architecture for AI agents.