Blog: Learn Machine Learning from a Google AI Engineer

I started my career as a software engineer over 20 years ago, and I’ve spent the last decade pioneering multimodal models and writing that viral prompt engineering paper you might have seen. As a Software Engineer at Google working within Google’s Office of the CTO (OCTO), my mandate is to design and innovate the next generation of Agentic AI systems and architectures that eventually shape Google’s global product and thought portfolio. Collaborating with engineers from Silicon Valley to our engineering hubs in Amsterdam and London, our team has pioneered real-world breakthroughs—including the operational speech agents behind the Wendy’s drive-thru success, which is now rolled out across America.

Today, our team is one of Google’s first “hybrid teams”, where humans and headless agents collaborate on real-world operational work. We are moving beyond treating AI as a simple autocomplete tool; we are embracing a co-worker model where AI agents take care of the heavy, time-consuming operational tasks, freeing our human minds for creative, strategic, and high-level architectural work.

But what nobody tells you is how quickly culture, process, and even risk management have to evolve when agents start building agents.