AI Systems
The modern AI stack: how Transformers turn text into predictions (built step by step on a live BERT), and the systems engineering — retrieval, RAG, agents — that wraps LLMs into products.
Two pillars share this domain. Generative AI builds the Transformer from first principles — tokens, embeddings, attention, the full encoder block — and ends at a live BERT pipeline you can probe sentence by sentence. Agentic AI & AI Systems then takes a trained model as given and asks the engineering question: how do you ground it in your own documents with retrieval and RAG, and how do you let it act — tools, multi-step workflows, and the failure modes that come with autonomy?
Nothing here is magic, but all of it leans on the vectors, probabilities and neural-network training told earlier in the catalogue. The stack below is the honest dependency order: each layer only works because the one beneath it does.