Amir Ivry

I am an incoming Postdoctoral Researcher at Carnegie Mellon University (Language Technologies Institute), working with Prof. Shinji Watanabe.

I study how to make spoken dialogue systems and voice agents trustworthy across three pillars: evaluation, safety, and security. My work develops mathematically principled and perceptually grounded evaluation methods to compare systems reliably under realistic conditions. I also investigate safety guardrails and security threats in human–agent voice interactions, including misuse and manipulation.

From 2015–2025, I worked in industry as a research scientist, most recently as Principal Researcher and Head of Data for Microsoft Teams. I completed a direct-track PhD at the Technion (2018–2023) on efficient personalization in speech systems, and was extremely fortunate to be advised by Prof. Israel Cohen and Dr. Baruch Berdugo.