In This Story
Keren Zhou’s relationship with NVIDIA spans more than a decade, evolving alongside the rapid rise of AI itself. What began as an early academic collaboration has become a sustained partnership centered on advancing the performance and efficiency of AI systems—work that now directly supports research and students at George Mason University.
Zhou, an assistant professor in the Department of Computer Science, traces his connection to NVIDIA back to his graduate studies. “I started collaboration officially with Nvidia about 12 years ago,” he said, describing a period when AI was in its infancy. “The only thing it could do is be used for image recognition…Is it a cat? A dog? That’s it.” At the time, his focus was not on AI applications themselves, but on the systems that make them possible. “I was never a native AI guy. What I was really thinking is, how do we make this system more efficient.”
That focus aligned naturally with NVIDIA’s strengths. Known for designing specialized chips that power modern AI, the company plays a central role in enabling the field’s rapid growth. Zhou, emphasizing that connection, said, “The theories behind AI was proposed many, many years ago, but it only works because of NVIDIA’s powerful chips, designed specifically and tailored for AI workloads.”
Over the years, Zhou’s collaboration with NVIDIA expanded, covering time across multiple institutions, including two separate internships he had with the company. Along the way, NVIDIA provided access to computing resources that supported his research into optimizing performance for scientific computing and AI workloads.
That relationship continues today at George Mason. Zhou recently received access to NVIDIA’s powerful DGX B200, a unified AI platform equipped with eight advanced graphics processing units (GPUs), which he uses to push forward his research. “We want to use it to train large models that can automatically generate efficient code,” he said. “Not only use like useful code, but also efficient code for AI systems.” The machine enables students to experiment with large-scale models and cutting-edge infrastructure that would otherwise be difficult to access.
Zhou’s work is consistently working toward improving the underlying systems that make AI possible. Even as his research incorporates AI more directly, that systems-oriented perspective remains central. “Even like until two years ago, what I was thinking about is just how to make AI run faster and more efficient and more robust.”
Now, with NVIDIA’s latest technology in place at Mason, that work continues with growing reach.