LLM Quantization and Inference

The questions GenAI interviewers actually ask — with handwritten explanations, diagrams and code for every answer.

End-to-end retrieval-augmented generation: chunking strategies, embeddings, hybrid search, reranking and evaluation — with full pipeline diagrams.

How large language models really work: tokenisation, pretraining objectives, instruction tuning, RLHF and decoding strategies, explained visually.

Everything a working developer needs in Python for AI: environments, NumPy, data pipelines, notebook workflows and the libraries the AI stack actually uses.

Tool use, planning loops, memory and multi-agent orchestration — a developer's field guide to building reliable agentic systems.

The transformer block, self-attention and multi-head attention drawn and annotated page by page — the clearest way to internalise the architecture behind modern LLMs.