AI Engineering Terms, Explained

The building blocks of modern AI systems, one at a time. Explained concretely, by the people who build them.

AI systems are made of a handful of building blocks that keep coming back: Token, Embeddings, Context Window, RAG, Evals. Understanding these terms is understanding why an AI project holds up or falls over.

This series explains them one by one. Each post takes a single term, starts from a concrete example rather than a definition, and stays easy to read. For more depth, every post has a separate, skippable technical box with the details.

The Machine

How a model runs, what it costs, and where the memory runs out first.

  • Token live
  • Context Window soon
  • Embedding soon
  • Attention soon
  • KV Cache soon
  • Quantization soon

The Harness

How a text generator turns into a system you can hand a task to.

  • Structured Output soon
  • Tool Use soon
  • Agents soon
  • RAG soon

The Discipline

How to measure, observe, and safely run AI systems for many tenants.

  • Evals soon
  • Observability soon
  • Prompt Injection soon
  • Multi-Tenancy soon

The Judgment

Which tool for which problem. The decisions that make a system.

  • Fine-tuning vs. RAG soon
  • Build vs. Buy soon

The series grows post by post. Terms without a link are in the works.

AI in your own company, not just as a term?

In 30 minutes we figure out which use case has the biggest leverage for you. Free, no commitment.

Book a conversation