Generative AI and LLMs

How large language models work, and how to build with embeddings, RAG and model APIs.

Comes after: Prompt Engineering

1. What Is Generative AI?Understand what makes generative models different, what they can produce and where they fall short.
2. How Large Language Models WorkFollow a prompt through tokenisation, next-token prediction and sampling, and see how a model is trained.
3. Embeddings and Vector SearchRepresent text as vectors of meaning and find the most relevant passages for a question, the basis of semantic search and RAG.
4. Retrieval-Augmented Generation (RAG)Give a model your own documents at question time so that it answers from facts, with sources, not from memory.
5. Working with Model APIsCall a language model from your own code: messages, parameters, streaming, cost and the errors you must handle.
6. Customising Models: Prompting, RAG or Fine-TuningChoose the right way to adapt a model to your needs, and understand what fine-tuning and LoRA actually change.
7. Image, Audio and Multimodal ModelsLearn how diffusion models create images from noise, and how models that see and hear are used in practice.
8. Evaluating Generative AIMeasure whether an application is accurate and grounded, with test sets, automatic checks and human review.
9. Responsible Generative AIRecognise the main risks of generative AI (bias, privacy, security, copyright and misuse) and the practices that manage them.
10. Project: Build a Document Question-Answering AssistantDesign and build a RAG assistant end to end: ingestion, retrieval, prompt, evaluation and safeguards.