Overview

AI engineering is the discipline of building AI systems from first principles — deriving algorithms from mathematics before introducing any framework. This contrasts with the prevalent tutorial-first approach of learning ML through frameworks (PyTorch, TensorFlow) without understanding underlying foundations.

The field spans from linear algebra foundations through neural network architectures, tokenization, attention mechanisms, and ultimately autonomous agent systems.

Math-first methodology

The core pedagogical principle: every algorithm is derived from raw math before a single framework is imported. Each lesson follows the pattern:

  1. Read the problem
  2. Derive the math
  3. Write the code
  4. Run the test
  5. Keep the artifact (runnable, multi-language implementation)

This approach is contrasted against fragmented AI education (scattered tutorials, framework-first shortcuts) to produce engineers who understand why algorithms work, not just how to call them.

Core topics

Open source curricula

AI Engineering from Scratch

Relationship to AI tooling

AI engineering is a prerequisite for understanding and effectively using higher-level tools like AI assistants, LLM wikis, and Claude Code. Understanding backpropagation and attention mechanisms from scratch enables informed decisions about model selection, fine-tuning, and system architecture.

Resources