This course explores the foundations of vectors and embeddings for representing text numerically and building semantic search systems. It focuses on the progression from preparing raw text and creating simple numerical representations to generating dense embeddings, measuring vector similarity, and retrieving semantically relevant information.

Through structured lessons and practical demonstrations, you will learn how text is cleaned and tokenised, how Bag-of-Words converts language into numerical vectors, and how vectors represent features in multidimensional spaces. You will also explore public embedding models on Hugging Face, generate embeddings with Sentence Transformers, batch-encode product data, and save, reload, retrieve, and verify embeddings efficiently. The course progresses from fundamental text representations to practical embedding-based search, emphasizing embedding spaces, model selection, similarity measures, normalisation, and reproducible workflows. Rather than treating embeddings as abstract numerical outputs, it focuses on understanding how they represent meaning, how they can be compared, and how they support semantic retrieval across a product catalogue. By the end of this course, you will be able to: - Prepare and tokenise text data for numerical representation - Build simple Bag-of-Words and NumPy vector representations - Explore and generate dense embeddings using Sentence Transformers - Batch-encode, validate, save, reload, and retrieve product embeddings - Compare vectors using cosine similarity, dot product, and Euclidean distance - Apply vector normalisation and build a brute-force semantic search workflow This course is ideal for aspiring AI engineers, machine learning practitioners, developers, data professionals, and learners interested in embeddings and semantic search. A foundational understanding of Python is recommended; prior experience with embeddings, vector search, or advanced retrieval techniques is not required. Join us to learn how to transform text into meaningful vector representations, generate and compare embeddings, and build the foundations of an effective semantic search system.