Cloud-Based Agentic AI Design Patterns
56mIntermediate2025-09-22
Authors
Thomas Erl
Course details
How should agents be deployed in the cloud? How should agents with multiple LLMs be designed? What are the best ways to optimize state data sharing, workload balancing, and solution performance? And, how can agentic solutions be designed to be resilient when having to interact with external services and tools that can be unreliable? These, and many other topics, are covered by LinkedIn Top Voice and best-selling author Thomas Erl, in this course featuring a diverse collection of design patterns geared toward agentic AI solutions in clouds, as well as on-premises environments. In addition to a range of design-centric techniques, this course also outlines agent management and observability practices, including agent activity tracing, policy enforcement, and LLM cost optimizations.
Skills covered
Software DesignAI Agents and Agentic SystemsBuilding with AICloud ServicesCloud ComputingSoftware DevelopmentOne-Off
Concepts
Introduction
- Introduction
- What you need to know
Agent Deployment and Functional Design Patterns
- Agent deployment and functional design patterns
- Containerized agent deployment
- Serverless agent capability functions
- Managed LLM endpoint
- Persistent agent activity state
- Materialized agent activity state view
Agent Scalability and Resilience Design Patterns
- Agent scalability and resilience design patterns
- Autoscaling agent fleet
- Load-balanced agent gateway
- External interaction circuit breaker
- Retry and idempotency
- Automated dead letter queue remediation
Agent Management and Observability Design Patterns
- Agent management and observability design patterns
- Automated externalized agent configuration
- Automated agent activity tracing
- Automated LLM cloud cost optimization
- Centralized automated agent policy enforcement
Conclusion
- Next steps