AI Engineer Guide · 2026 edition
Learn AI engineering as a connected system.
From tokenization and attention to RAG, agents, agentic coding and production failure modes. Every concept has a page; the first mention in each section links to it with an instant definition.
- Concepts
- 460
- Published
- 460
- Questions
- 76
- Tracks
- 19
The curriculum
All tracks →- 00OrientationWhat AI engineering is, and how it differs from ML and research.12 pages
- 01Software foundationsThe engineering base: APIs, streaming, queues, testing, deployment.24 pages
- 02Math, data & ML foundationsVectors, probability, training, metrics, and the statistics interviews probe.27 pages
- 03Deep learning & transformersAttention, tokenization, inference mechanics, and the transformer block.33 pages
- 04LLM applicationsTokens, decoding, structured output, tool calls, and cost.28 pages
- 05Prompting & context engineeringInstructions, context assembly, and defending untrusted input.26 pages
- 06Embeddings, search & RAGFrom parsing and chunking to hybrid retrieval, reranking and citations.57 pages
- 07Agents & orchestrationLoops, tools, memory, approvals, durable execution, and when not to use an agent.37 pages
- 08Agentic codingRepository context, patch editing, verification, worktrees and sandboxing.30 pages
- 09Tool protocolsMCP, A2A, and the difference between capability access and delegation.16 pages
- 10Model adaptationSFT, LoRA, preference optimisation, and when tuning is the wrong tool.22 pages
- 11Multimodal, vision & voiceVision, speech, real-time voice, and document understanding.22 pages
- 12Evaluation & qualityGolden sets, judges, regression gates, and error analysis.28 pages
- 13Security, privacy & governanceInjection, exfiltration, excessive agency, and layered controls.0 pages
- 14Production & LLMOpsLatency, cost, observability, failover, and incident response.38 pages
- 15System designReference architectures and twelve end-to-end design walkthroughs.0 pages
- 16Data, SQL & storageSQL, modelling, pipelines, and vector stores next to relational data.17 pages
- 17Career & interview craftAnswer frameworks, drills, take-homes, and behavioural preparation.33 pages
- 18Frameworks & orchestrationLangGraph, LangChain, LlamaIndex, CrewAI, SDKs, and how to choose.0 pages