LLM (Large Language Model) has evolved into a vogue (and by implication vaguer) term, with an ever-growing portfolio of tools, applications, parameters, architectures, and AI / ML approaches. It’s meaning has extended so far and wide that sometimes LLM seems to be the only type of AI in the market, or useful to organisations. Such proliferation makes it rather challenging, especially for non-specialists, to understand the differences between all the “LLMs” out there and what each “LLM” can do well or not. This course will help the audience to separate marketing talk, captivating headlines, and typical glossing (so-called “typicality bias”) of the technology from more dependable facts. It provides critical clarity for decision making around this technology in terms of procuring, governing, deploying, and relying on LLMs directly (e.g. Natural Language Processing, NLP), with support (e.g. via Retrieval Augmented Generation, RAG), or via extensions (e.g. chatbots and Agentic AI). This course also provides the foundation to assess and better manage typical risks, like vendor lock-in.

By looking under the hood just enough, the audience will be on a more practical and successful trajectory for many AI projects, because as even Gemini AI itself summarised it in a recent real search result: While sometimes simplified as just "predicting the next word," in practice,...

This course extends our Deep Learning Essentials course, and complements the Agentic AI Primer course.

Who should attend?

• C-Suite and Board members, DPOs, and Compliance Professionals
• Privacy / Software Engineers, Technical Staff, Developers, Data Analysts, Data Architects, and Project / Risk Managers
• IT Practitioners and engineers tasked to deploy LLM based technologies
• Executives, Managers, and Staff involved in the management, collection, use or other processing of personal data

Course Details

Course Details

• Key terms for Natural Language Processing (NLP)​
  • Natural Language Understanding (NLU), Natural Language Generation (NLG)
  • Artificial Intelligence (AI), Machine Learning (ML)
  • Traditional NLP approaches, LLM for language
• Basic Transformer and LLM architectures
  • Encoder
  • Decoder
  • Encoder-Decoder
  • Autoregressive
  • Probabilistic / deterministic versus GenAI
  • Embedding, Attention, ‘next word prediction’
  • LLM versus ChatBot versus Agentic
• Enhancements for LLMS 
  • Chain-of-Thought  and ReAct
  • Retrieval Augmented Generation (RAG)
  • Fine-Tuning
  • Token tuning
  • Reasoning Models

Course Facilitator

Albert Pichlmaier  Senior Cybersecurity and Privacy Engineer Senior Learning Technology Designer, Drew Data Protection & Cybersecurity Academy