Where AI Systems Break in Production

​This talk focuses on how AI work actually moves through the system once it is under real load. What looks like a simple request at the surface often behaves very differently as it flows through different layers. Over time, certain patterns begin to dominate system behavior, especially when different types of work start interacting through shared resources. The session looks at how this shows up in practice and how to recognize it early enough to keep the system stable as it scales.

​Speaker: Bruno Samuel, Head of Platforms, Amalgus Global.

​Bruno works on taking early-stage technical ideas from foundational research to production at scale. He focuses on defining and building solutions for problems that are not well-specified at the start, and ensuring they hold up in real-world environments. He has led engineering initiatives and teams across global systems at Meta, PayPal, eBay, and Everbridge.