AI Agents in Production
What changes when an agent loop moves from a demo to production: hard limits, tool authorisation, failure handling, and knowing when an agent is the wrong shape entirely.
intermediate· 4 minLLMAgents
Pillar
AI Engineering
How AI applications are actually built and run: retrieval architecture, agent orchestration, tool calling, structured outputs, evaluation, and the observability required to keep any of it trustworthy under load.
What this section covers
Consolidated guides rather than a page per concept. Each area below is one resource, or will be, rather than a cluster of near-duplicates.
Application architecture
How the pieces assemble: request path, retrieval, model call, tool execution, and where to put boundaries so failures stay contained instead of cascading.
Retrieval and context
Chunking, embeddings, ranking, and context budgets — the layer where most "the model got it wrong" bugs actually live.
Agents and tool calling
Loop control, iteration limits, cost budgets, and treating the tool boundary as a real authorisation boundary rather than a polite prompt instruction.
Evaluation and observability
Measuring non-deterministic systems: graded suites, tracing across spans, cost attribution, and catching a regression before a user reports it.
Resources
Resources in this pillar
Every resource states its author, its review status, and whether the procedures in it were executed or only reviewed.
Production AI Application Architecture
How to structure an AI application so that a slow model, a failed tool call, or a bad retrieval degrades one part of the system instead of the whole request path.
intermediate· 6 minLLMRAG
RAG Architecture Guide
A working guide to retrieval-augmented generation: how to chunk, how to evaluate retrieval separately from generation, and how to diagnose a bad answer in the right order.
intermediate· 5 minRAGEmbeddings
The AI Stack Explained
A layer-by-layer map of the modern AI stack, what each layer is actually responsible for, and where production systems tend to break in practice.
foundational· 8 minLLMKubernetes
Where this connects
No layer of the stack is operated in isolation. These are the sections you are most likely to need next.
AI InfrastructureWhen inference is slow the answer is usually here: memory bandwidth, batching, and the fabric between accelerators.Go →AI OperationsDiagnostic method for systems whose output you cannot write an equality assertion against.Go →AcademyStructured paths that build these skills in sequence instead of by topic hopping.Go →
Newsletter
Inside The AI Stack Brief
A practical weekly briefing on AI engineering, infrastructure, production operations, and the technologies powering the AI stack.
One email a week. No sponsorship placements inside the technical sections. Unsubscribe in one click.