The chapters — one page each
- 01Learning from dataBefore any neural network: what it means for a machine to learn, and the discipline that keeps you from fooling yourself.
- 02The classical toolboxMost production ML is not deep: trees, ensembles and linear models carry tabular data, and knowing them keeps you honest.
- 03Deep learning fundamentalsStack linear layers with nonlinearities and train by gradient descent; the questions become why it works, and which knobs matter.
- 04What the model actually doesEverything downstream gets easier once the model stops being a black box.
- 05Serving and inference economicsBetween the model and your users sits a serving stack with its own physics; its numbers set your product’s.
- 06Prompting and structured outputThe lowest rung of the leverage ladder — and the one that solves more problems than its reputation suggests.
- 07Context and retrievalThe model is fixed; what you control is what it reads. Retrieval is how you control it well.
- 08Customisation: fine-tuning and when not toThe escalation ladder’s top rung — expensive, powerful, and reached far less often than assumed.
- 09Agents and toolsGive the model the ability to act, and a new discipline appears: deciding when it should.
- 10Security: injection, leakage and sandboxingThe model reads everything and believes most of it; the systems around it must not.
- 11Evaluation and productionThe demo works. This stage is the difference between a demo and a system.
- 12In production: feedback, drift and improvement loopsOffline evals said ship it; now the system meets users, and the measuring must continue.