Language is the data type that changed everything. Models that can read, summarise, translate, and write have moved from research labs into everyday tools, and understanding how they work is now a core skill for any data scientist.
You'll start with a big-picture tour of large language models, then dig into the architecture behind them: the transformer and its attention mechanism. From there you'll see how raw text becomes numbers through tokenization, and finish by building a small GPT from scratch so the whole pipeline, from text in to text out, makes sense end to end.
With these ideas in place, you'll be ready for the final module, MLOps & Deployment, where you'll learn to ship models and keep them running reliably in production.
A one-hour overview of what LLMs are, how they are trained, where they are heading, and the security challenges they bring.
A visual walk through the transformer: embeddings, how text is predicted one token at a time, and what happens inside the model.
How queries, keys, and values let each word look at its context, building up to full multi-headed attention.
How text is split into tokens with byte-pair encoding, and why tokenization explains many odd LLM behaviours.
Build and train a small GPT-style transformer step by step in code, from a bigram baseline to self-attention.