Two compendiums, free to download and reuse.

The machine intelligence stack from the mathematics underneath it to the systems built on top of it: linear algebra, probability and information theory; neural network mechanics; the transformer in detail; pre-training, fine-tuning and alignment; reinforcement learning; inference, serving and accelerators; retrieval, prompting and agents; reasoning models, evaluation and safety. Every chapter ends with review questions, and the studies in depth take one system or failure mode at a time.

1,164 concepts · 12,029 review questions · 145 studies in depth · 101 tracks

Prefer to read on the web? The living edition is at praveentn.live/learn.

The estate a solution architect is accountable for: the fundamentals and patterns of the discipline, distributed systems and the data they disagree about, cloud, networking, security, reliability, operations, cost and governance. Every topic gives its defining entries, scenario studies that resolve into a decision, and a rapid-review set. The book closes with worked architectures reproduced as delivered and field guides built from published evidence.

600 topics · 2,513 entries · 3,294 scenario studies · 6,558 review questions · 55 deliverables

Prefer to read on the web? The living edition is at praveentn.live/architecture.