Everyone is using large language models. Far fewer people understand what is actually happening inside them. This masterclass changes that. In one intensive half-day session, you will go beyond the surface-level familiarity with tools such as ChatGPT or Mistral and develop a rigorous understanding of the architecture and training processes that make these systems work, and why they sometimes do not.

Whether you are a non-technical leader wanting to make more informed strategic decisions, or a technical professional looking for a solid conceptual foundation, this session bridges the gap between high-level business utility and the underlying mechanics. You will leave knowing not just what LLMs can do, but how they do it, where they might fall short, and how to bring them into your organization responsibly and cost-effectively.

What you will learn

The masterclass is structured around four core themes.

How LLMs work under the hood. You will demystify the transformer architecture, understanding how attention mechanisms allow models to process context and meaning, and how tokenization and embeddings form the "language" that machines use to interpret human text. You will also trace the full lifecycle of an LLM, from large-scale pre-training on raw data through to supervised fine-tuning and reinforcement learning from human feedback (RLHF).

Situating LLMs in the broader AI landscape. You will understand how LLMs fit within the wider fields of deep learning and natural language processing, and how generative AI differs from classical predictive systems.

Limitations and critical assessment. You will learn to critically assess the real limitations of LLMs, including stochasticity, context limits, and bias, so you can set realistic expectations for any project involving these systems.

Responsible adoption. You will apply a Buy vs Build vs Tune framework to determine the most cost-effective and secure way to integrate LLMs into your organization, and evaluate the data privacy and ethical risks involved, particularly around the handling of sensitive corporate information via third-party APIs. You will also get an overview of the current state of the art in 2026.

Who is this masterclass for?

This masterclass is aimed at SMEs, startups, and independent professionals. No technical background is required. If you want to truly understand how LLMs work, not just use it, and build the critical foundation needed to engage with LLMs confidently in your organization, this masterclass is for you.

Price 

Thanks to the support of the European Commission and Innoviris in the framework of the EDIH sustAIn.brussels, SMEs and midcaps receive this training free of charge in the context of de minimis aid.

Practical information

• Language: English (bilingual exchanges in French and English are welcome)
• Location: BeCentral, Cantersteen 12, 1000 Brussels
•  Format: in person, interactive, hands-on
•  Participants: maximum 5
• Duration: 4 hours

Your trainer

Marjon Blondeel is an AI-engineer and professor at the Vrije Universiteit Brussel (VUB) AI Lab. She holds a PhD in Artificial Intelligence from VUB and Ghent University, as well as a Postgraduate Certificate in Education. At the lab, she develops and maintains AI demos in collaboration with researchers and leads workshops for audiences ranging from high school students to business professionals. These sessions cover a broad spectrum from introductory AI concepts to technical programming. She also lectures in the Computer Science and Artificial Intelligence bachelor programmes at VUB.

Questions?

Contact Marco Houben at marco.houben@vub.be