Agentic AI: Actor Models and Subagent Architecture

About this course

Learn to build production agentic AI systems using actor model foundations, subagent architecture patterns, and multi-language implementations. You will explore the actor paradigm for concurrent computation, where isolated processes communicate through message-passing with zero shared memory, eliminating race conditions and deadlocks that crash production systems. The course covers Actix supervision trees in Rust for fault-tolerant actor recovery and location transparency for seamless distributed scaling. You will implement Claude subagent patterns for task-specific AI configurations with isolated state and tool access, and examine pmat subagent architecture for code quality analysis through specialized delegation pipelines. The subagent module demonstrates supervised multi-agent coordination, applies Amdahl's law to understand parallelization limits of subagent systems, and explains why simple agents often outperform complex multi-agent designs. You will also explore small language models as efficient alternatives for agent reasoning tasks. The hands-on module covers actor implementations in three languages: Deno with TypeScript, Go with goroutines and channels, and Rust with ownership-based memory safety. You will build Go supervisor patterns for automatic actor recovery and examine a complete agentic coding project repository. By completing this course, you will be able to design fault-tolerant agentic systems using actor model principles, implement subagent architectures with Claude, and build actor patterns across multiple programming languages.

Prices, discounts and availability are set by Coursera. We may earn a commission when you purchase through links on this site.