To Bill Dally, few fields felt as dynamic and transformative as computing in the mid-1980s. “The technology was improving so rapidly that new things became possible every year that hadn’t been possible just a few years before,” he says.

So, when he went looking for a PhD program that would connect him with some of the most important ideas in the field, he chose Caltech. Dally was intrigued by the pioneering contributions of Carver Mead (BS ’56, MS ’57, PhD ’60) to very large-scale integration (VSLI) technology, a process that greatly increased the number of transistors that could be placed on a semiconductor chip.

At Caltech, Dally worked with Charles Seitz on multiprocessors. After earning his degree, he joined the faculty at MIT, where he focused on fundamental issues linked to parallel computing and interconnection networks. He and his team built experimental parallel computer systems that demonstrated fast communication and synchronization mechanisms. Dally also created much of the underlying theory for interconnection networks.

Later, at Stanford, he developed stream processing, which led to graphics processing unit (GPU) computing. He also constructed accelerators that enhanced the performance of deep learning and bioinformatics applications, including the rapid assembly of genomes.

Even though Dally’s academic pursuits centered on fundamental theory, he often had an eye to practical applications. Together with classmates, he founded his first start-up while still at Caltech, and he cofounded three others before landing at the technology company NVIDIA.

At NVIDIA, Dally has helped pioneer advances in deep-learning platforms that support everything from image analysis to language translation to self-driving cars. “Here, you get to see ideas go from theory to prototypes in the lab to real products for real people,” he says.

He sees enormous possibilities ahead. “You get to make big jumps,” he says. “There’s always a new game.”