Course · 1 modules · 12 lessons · 112 min
Agentic Design Patterns
Architecture selection, tool design, error resilience, multi-agent coordination, and production patterns for agentic systems.
The course at a glance
1 module · click any tile to jump to its lessons.
All lessons
№ 01Design Patterns
01Agent Testing StrategyA comprehensive testing architecture for agent systems, covering the testing pyramid, non-determinism management, evaluation design, and CI/CD pipelines.→02Architecture Selection FrameworkA decision matrix for choosing the right agent architecture pattern based on your task's structure, latency requirements, cost constraints, and error tolerance.→03Complexity GradientStart with the simplest LLM pattern that could work and add complexity only when measured evidence proves it insufficient.→04Context and State StrategySystematic approaches for allocating, prioritizing, compressing, and persisting information within and beyond the context window.→05Cost-Latency OptimizationDesign strategies for building agent architectures that meet cost and latency targets without sacrificing task quality.→06Do You Need an Agent?A decision framework for determining whether your problem requires an autonomous agent or whether a simpler alternative will perform better at lower cost.→07Error Resilience PatternsArchitectural patterns for designing agent systems that anticipate, contain, and recover from failures before they happen.→08Incremental Migration PatternsPatterns for introducing agent capabilities into existing systems gradually, minimizing risk while capturing value at each stage.→09Multi-Agent Decision FrameworkA decision framework for determining when multi-agent architectures are justified and how to design agent boundaries, communication, and coordination when they are.→10Safety by DesignA holistic approach to engineering safety into agent architecture from the ground up, covering threat modeling, layered defenses, permission design, and the safety-capability tradeoff.→11System Prompt EngineeringThe practical craft of writing system prompts that reliably steer agent behavior, manage tool use, and enforce constraints within token budgets.→12Tool Interface DesignDesign principles for building tool APIs that LLMs can discover, understand, and invoke correctly without human intervention.→