Nice to meet you!
I'm Juliette Pluto, a Staff Software Engineer at Google DeepMind in New York City. I work on adversarial robustness for frontier AI systems. Lately that has mostly meant prompt injection: what an agent should do when the data it reads is trying to give it orders.
I started Google DeepMind's first effort to defend against prompt injection. Since then the work has ranged from the security evaluation bundle used in frontier model selection, to robustness work on Gemini, to an early policy-based defense whose approach was later adopted to harden production safety systems against human red-teaming. The common thread: turning messy AI security risks into evals, evidence, and practical defenses.
I got here by an indirect route: helping scale N26's web platform as it became a unicorn, then trusted execution environments and privacy-preserving ML in my first years at Google, then this. I tend to move quickly through unfamiliar technical terrain, and I've come to like it that way. The details are in my CV.