There are two types of engineers working on AI today: those building agents, and those who think they can’t.

The funny part? The second group is often better prepared for it.

Before LLMs, most machine learning work stayed in research. Scale, reliability, and long-term maintenance weren’t the main concern. With LLMs, that changed. We no longer train models as much, we build systems around them.

And that’s where the gap shows.

Many agentic applications are built by ML engineers now facing production challenges for the first time. Meanwhile, software engineers, who already know how to design scalable and reliable systems, often hesitate to step in.

But what if this isn’t really an AI problem?

What if it’s a distributed systems problem?

In this talk, we’ll explore how classic system design principles apply directly to agentic applications, covering state, orchestration, reliability, and observability.

If you’ve built microservices or distributed systems, this will feel familiar.

When we’re done, you’ll see that AI engineering isn’t something new, it’s something you already know how to do.