In this demo-driven engineering session, we will explore the hidden aspects of Generative Artificial Intelligence that go far beyond prompt engineering and basics. The goal of the session is to learn how GenAI works and how to implement various use cases, such as classification, recommendation, and semantic search.

We will start with core topics such as tokens, demonstrating how transformers learn associations between them. Next, we will discuss completions to explain how text generation works, providing a deeper understanding of model functionality. Additionally, we will cover embedding vectors, cosine similarity, and vector databases to explain the semantics behind models.

Afterwards, you will learn how to expand and manipulate model knowledge using techniques like Retrieval Augmented Generation (RAG).

Finally, we will explore different model types and tools such as OpenAI, Microsoft Foundry (also a local one) and more.

This session is a technical presentation designed for anyone interested in understanding and building intelligent applications leveraging large language models.