Learning path
How LLMs Work
Tokens, embeddings, attention, temperature and hallucinations: how large language models really work, with runnable Python. Part of the Gen AI Developer premium track.
-
1
Tokens and Next-Token Prediction
A large language model does one thing: predict the next token. Build a tiny next-word predictor in plain Python and see how text generation really works.
-
2
Embeddings: Meaning as Numbers
Embeddings turn words and sentences into lists of numbers where similar meanings sit close together. Compute cosine similarity by hand and build a mini semantic search.
-
3
Transformers and Attention, Visually
Attention lets each word look at the other words that matter for it. Compute attention weights with softmax in plain Python and see why transformers understand context.
-
4
Transformers and Attention, Visually
Attention lets each word look at the other words that matter for it. Compute attention weights with softmax in plain Python and see why transformers understand context.
-
5
Context Windows, Temperature and Sampling
Why the same prompt gives different answers, what temperature and top-k really do, and why models "forget" in long chats. With runnable Python you can tweak.
-
6
Why Models Hallucinate and How to Reduce It
LLMs sometimes state false things with full confidence. Learn why, and five practical defences, including grounding answers in sources and checking them with code.