AI Wisdom
Learning Track

AI Engineering

RAG, multi-agent systems, prompt engineering, token economics, evaluations and guardrails — the production discipline of building with LLMs.

12+
Topics
11
Published
100%
Free
Browse articles ↓

12 articlesin AI Engineering

AI Engineering
EmployeeHelpdesk appModelToolsOrders, accountsRun traceApproverTests, schema

Agent Design Patterns: Start One Rung Lower

Advanced

Prompt chaining, routing, orchestrator-workers, ReAct, plan-and-execute, reflection and tree search: what each costs, what the research shows, and when to climb to the next one.

8 min
Read →
AI Engineering
EmployeesAssistantInference serverKV cacheModel on your GPUPrefix cacheDraft model

Model Serving With vLLM: The KV Cache Decides Your Capacity

Advanced

KV cache arithmetic, PagedAttention, continuous batching, prefix caching, FP8 and speculative decoding, the vLLM flags that matter, and the five metrics that show which limit you hit.

8 min
Read →
AI Engineering
CallerSpeech to textAgentText to speechParcel lookupA person

Voice Agents: A Real-Time Problem With a Model Inside

Advanced

Speech-to-speech or a pipeline, a latency budget for every stage, interruptions that really work, prompts written for the ear, and what a call actually costs.

Updated 5 min
Read →
AI Engineering
UsersInput checksLLMTool policyRefund toolOutput checks

Guardrails: Input/Output Safety in Production

Intermediate

Input, output, schema, tool, and cost guardrails — the layered defence system that catches LLM failures before they reach a user or downstream system.

Updated 13 min
Read →
AI Engineering
UsersSearchEmbedding modelVector indexNew index

Embeddings: From Theory to Production Choice

Intermediate

Convert text, images, and code into dense vectors that capture semantic meaning — the foundational primitive of modern AI systems, and the choice that dominates RAG quality.

Updated 11 min
Read →
AI Engineering
A production AI system on Azure: an orchestrator classifies intent with a cheap model, routes to RAG over Azure AI Search with hybrid retrieval or to direct handlers, enforces token budgets, and traces every step.A production AI system on Azure: an orchestrator classifies intent with a cheap model, routes to RAG over Azure AI Search with hybrid retrieval or to direct handlers, enforces token budgets, and traces every step.

How to Build a Production-Ready AI System (Azure OpenAI + AI Search — Real Architecture)

AdvancedPRO

Azure OpenAI + AI Search + embeddings — real-world architecture for production AI systems, including legacy data, orchestration, hybrid retrieval, cost control, and failure modes.

Updated 15 min
Read →
AI Engineering
UsersChatbotLarge modelToken logPrompt cacheSmall model

Token Economics: Understanding and Optimizing LLM Costs

A practical guide to understanding token pricing, measuring real costs, and implementing optimization strategies — caching, prompt compression, model routing.

Updated 10 min
Read →
AI Engineering
EmployeeHR portalKernelModelLeave pluginHR servicesTraces

Building Reliable AI Agents with Semantic Kernel

Intermediate

Plugin architecture, memory, planners, and error handling patterns for building production AI agents in .NET with Semantic Kernel.

Updated 11 min
Read →
AI Engineering
A personal knowledge base: notes from Notion, Obsidian, OneNote and files are chunked at 300 to 500 tokens, embedded with text-embedding-3-small into a local vector store, and queried through a chat interface with sources.A personal knowledge base: notes from Notion, Obsidian, OneNote and files are chunked at 300 to 500 tokens, embedded with text-embedding-3-small into a local vector store, and queried through a chat interface with sources.

Building a Personal AI Knowledge Base

How to build a personal RAG system over your notes, bookmarks, and documents — using embeddings, vector search, and a conversational interface.

Updated 11 min
Read →
AI Engineering
TicketsTicket routerCategory guideTuned small modelHeld-out test set

When to Fine-Tune vs Few-Shot vs RAG

Intermediate

A decision framework for choosing between fine-tuning, few-shot prompting, and RAG for production LLM applications.

Updated 10 min
Read →
AI Engineering
Twelve microservices call Azure OpenAI through one AI gateway on Azure API Management that applies token-based rate limits, semantic caching and per-team cost tracking, and fails over between regional deployments.Twelve microservices call Azure OpenAI through one AI gateway on Azure API Management that applies token-based rate limits, semantic caching and per-team cost tracking, and fails over between regional deployments.

The AI Gateway Pattern: Why Every Production LLM Needs One

Intermediate

API Management, rate limiting, semantic caching, and cost control with Azure APIM as an AI Gateway.

Updated 10 min
Read →
AI Engineering
UsersAssistantVector indexKeyword indexRerankerLLM

Designing RAG Systems That Actually Scale

Intermediate

Chunking strategies, embedding pipelines, retrieval patterns, and when RAG breaks down in production systems.

Updated 12 min
Read →