Trudy Lu

About Trudy Lu

A chip design director focusing on heterogeneous computing and SoC architecture, currently leading the research and development of computing subsystems and customized acceleration modules under the multi-core ARM architecture at NVIDIA, committed to building AI-oriented, large-scale chip platforms. Through close cooperation with the architecture, packaging, software and verification teams, multiple high-performance platforms are promoted from concept to implementation, and the boundaries of silicon wafers are continuously expanded. I led the team to participate in the front-end design of NVIDIA Grace Hopper super chip, which significantly improved the performance and efficiency of large language model training. At the same time, we are also promoting Blackwell The platform\'s on-chip interconnect architecture provides higher computing density and bandwidth paths for the next generation of multi-chip modules. I am not only concerned about the chip itself, but also how computing changes the relationship between humans and the world. Inside the chip, there is a game of logic and charge; outside the chip, there is an expansion of cognitive boundaries. Besides work, I always maintain my curiosity and passion for exploring the world. I love island travel and diving, and enjoy observing the flow logic of natural systems underwater. I also like hiking in the mountains and forests of Northern California to experience that nature, like algorithms, has its boundaries and convergence. I also like philosophy and science fiction. In my eyes, there is no obvious boundary between Zhuangzi and Nabokov, Wittgenstein and \"The Three-Body Problem\"- they are all exploring the nature of the world and the boundaries of human beings. Sometimes I think about a question late at night: If the chip can think, how will it view the way we design it?

Experience

1. Director of Chip Design

   NVIDIA

   Jul 2021 — Present · San Jose, CA, US

   Responsible for SoC technical leadership, and the execution of an ARM based multi-core compute subsystem and accelerators: with focus on front-end architectural pathfinding, pre-si design & verification.

Education

• Boston University

  M.S