Traditional request/response architectures struggle to support LLM-powered autonomous agents — leading to broken workflows, delayed actions, and brittle pipelines. In this talk, we introduce the StreamNative Agent Engine, an event-driven framework evolved from Pulsar Functions, purpose-built for real-time, autonomous AI systems where every perception, plan, and action is represented as an event.

You’ll see how the Agent Engine transforms static agent pipelines into event-driven, resilient, and scalable runtimes, enabling continuous interaction, feedback, and coordination among multiple agents — without blocking the LLM or sacrificing reliability.

What we’ll cover:

• How declarative agent meshes are compiled to run as event-driven functions.
• Workflow and agent orchestration using streaming joins and time windows.
• Techniques for back-pressure handling and retries that keep LLMs responsive.
• Idempotency, observability, and safety patterns for real-time agent systems.
• Real-world lessons learned running autonomous agents at scale.

If you’re an AI engineer or platform architect building agentic systems that react, plan, and adapt in real time, this talk will show you how event-driven architectures can turn AI pipelines into self-governing, autonomous networks.