Skip to content
AI Engineer Guide.

Curriculum

19 tracks and 2 collections. Learning order is guidance, not a gate.

OrientationTrack · 12 pages

What AI engineering is, and how it differs from ML and research.

Software foundationsTrack · 24 pages

The engineering base: APIs, streaming, queues, testing, deployment.

Vectors, probability, training, metrics, and the statistics interviews probe.

Attention, tokenization, inference mechanics, and the transformer block.

LLM applicationsTrack · 28 pages

Tokens, decoding, structured output, tool calls, and cost.

Instructions, context assembly, and defending untrusted input.

Embeddings, search & RAGTrack · 57 pages

From parsing and chunking to hybrid retrieval, reranking and citations.

Agents & orchestrationTrack · 37 pages

Loops, tools, memory, approvals, durable execution, and when not to use an agent.

Agentic codingTrack · 30 pages

Repository context, patch editing, verification, worktrees and sandboxing.

Tool protocolsTrack · 16 pages

MCP, A2A, and the difference between capability access and delegation.

Model adaptationTrack · 22 pages

SFT, LoRA, preference optimisation, and when tuning is the wrong tool.

Vision, speech, real-time voice, and document understanding.

Evaluation & qualityTrack · 28 pages

Golden sets, judges, regression gates, and error analysis.

Injection, exfiltration, excessive agency, and layered controls.

Production & LLMOpsTrack · 38 pages

Latency, cost, observability, failover, and incident response.

System designTrack · 0 pages

Reference architectures and twelve end-to-end design walkthroughs.

Data, SQL & storageTrack · 17 pages

SQL, modelling, pipelines, and vector stores next to relational data.

Career & interview craftTrack · 33 pages

Answer frameworks, drills, take-homes, and behavioural preparation.

LangGraph, LangChain, LlamaIndex, CrewAI, SDKs, and how to choose.

ToolsCollection · 3 pages

Infrastructure comparisons: providers, vector stores, serving, observability.

Architecture patternsCollection · 7 pages

Reusable topologies: queues, fan-out, cache-aside, sidecars, event-driven evals.