Where to begin? This summer, GPT-5 made people nostalgic for its predecessor, in recent weeks the internet was flooded with disturbing Sora-generated videos, and just yesterday Sam Altman tweeted ChatGPT will soon allow erotica for adult users. Do you also feel like it's time to catch a breath? Then the next Belgium NLP Meetup is something for you.
On November 20, North Star and Faktion are inviting us to The Beacon in Antwerp. Talks begin around 7.30pm, and from 9pm onward you're welcome to stay for informal networking and conversation.
Below are the presentations of the first two speakers. Younes Baghor, CEO of BrainBlend AI, will present their popular Atomic-Agents framework, while Aleksandra M.W. Vercauteren, Head of AI at Greenomy, will show us how we can design reliable agentic flows.
From Monolith to Multi-Agent: Architecting Production-Ready AI Systems with Atomic-Agents
Younes Baghor, BrainBlend AI
Many developers struggle to scale and maintain their agentic AI prototypes built with frameworks like LangChain, which often evolve into complex monoliths. In this talk, Younes Baghor, CEO of BrainBlend AI, will do a technical deep dive into the core principles of BrainBlend's Atomic-Agents library, a modular, low-level framework designed for building robust and scalable multi-agent systems grounded in traditional software architecture principles. He will also introduce their “Blueprint-First” methodology: a process that uses “Process Friction Analysis” to align technical design with real business needs. You'll go home with a practical framework for turning AI concepts into production-ready, business-aligned systems.
Making Agents Behave in Data Structuring
Aleksandra M.W. Vercauteren, Greenomy
Designing reliable agentic flows for data structuring sounds simple, until you try. In this talk, Aleksandra will walk us through three different approaches she explored in her work for Greenomy: a ReAct-style agent, a Plan-Execute setup, and the final, homegrown approach that actually worked. Expect practical insights, some pitfalls, and a few counterintuitive lessons on making agentic systems behave.
