From Algorithms to Intelligence

Artificial intelligence is now something every engineering organisation is expected to understand, adopt and govern. The people asked to make that happen are usually experienced software engineers and architects who never studied machine learning. This book is written for them.

From Algorithms to Intelligence takes you from the first idea of a learning machine to building, adapting, evaluating and governing language models. Every important idea is introduced with a story, explained in plain English, worked through in mathematics by hand, implemented in Python code that was actually run, and then broken on purpose, because understanding how a system fails is the surest way to understand how it works.

Inside you will:

• Rebuild the mathematics of learning from vectors and derivatives to probability, with worked examples
• Build regression, classification, trees, ensembles and neural networks from scratch, then in scikit learn and PyTorch
• Construct tokenisers, embeddings, attention and a transformer, and train your own miniature language model on an ordinary laptop
• Adapt models with LoRA, evaluate them honestly, quantise them and size the hardware they need
• Design retrieval, agents, security, MLOps, governance and cost controls the way an architect must
• Follow one capstone, Archie the AI Architecture Assistant, as it grows from keyword rules to a tested, secured service

With 48 chapters, more than 180 runnable programs and all companion code on GitHub, this is a complete course for engineers and architects who want to understand AI from the inside and invent what comes next.