Robert Hundt, who previously led software development for Google's tensor processing unit (TPU) chips, has joined Amazon. His new role focuses on building the accompanying software that developers use to run AI models on Amazon's in-house Trainium chips.
An Amazon spokesperson confirmed that Hundt, a Google distinguished engineer and the initial software lead for Google's TPU chips, has officially joined Amazon's chip team with the same rank of distinguished engineer. The spokesperson stated that Hundt will report to Amazon Senior Vice President Peter DeSantis and will initially focus on Amazon's Neuron software system. This suite of tools allows developers to deploy and run various AI models on Amazon's proprietary Trainium chips—designed for AI training and inference—competing directly with Nvidia's widely used CUDA software ecosystem.
During his two-decade tenure at Google, Hundt was responsible for optimizing machine learning models to ensure algorithms ran efficiently on Google's TPU (used for AI training and inference) as well as third-party GPU and CPU hardware. He played a key role in creating XLA (Accelerated Linear Algebra Compiler), a software tool launched by Google in 2017 that serves as the foundational layer for hardware adaptation.
Three months ago, Amazon also hired another former Google employee, Steve Molloy, as vice president of AI chips. Molloy previously developed custom chips for Google's Pixel series phones and tablets. Hundt's addition further strengthens Amazon's chip software capabilities. The Amazon spokesperson added that in the long term, Hundt will also dedicate some of his efforts to helping the company build its overall quantum computing technology roadmap.
Comments