AI Engineering Patterns

Case Studies

Real-world applications of AI engineering patterns in production.

Claude / Anthropic

Claude – Safety-First Multi-Agent Research System

How Anthropic's Research feature uses an orchestrator-subagent architecture with MCP, context compaction, and Constitutional AI-derived trust boundaries to execute parallel, open-ended research at scale.

Multi-Agent OrchestrationContext CompactionEffort ScalingHuman-in-the-LoopSpan-Level Tracing

OpenAI Agents SDK

OpenAI Agents SDK – Primitive-First Agent Orchestration Framework

How OpenAI's Agents SDK delivers production-grade multi-agent orchestration through four core primitives — Agents, Handoffs, Guardrails, and Sessions — built on the Responses API.

Multi-Agent OrchestrationGuardrailsContext CompactionSpan-Level TracingCircuit Breaker for LLMs

Perplexity AI

Perplexity AI – Real-Time AI Search Engine

How Perplexity delivers sub-2s grounded answers at scale using hybrid retrieval, multi-model routing, and aggressive caching.

Hybrid SearchModel RouterSemantic CachingSpan-Level TracingToken Budget PatternRetrieval Freshness Watermark