Industry veteran Jay Boisseau – former director of the Texas Advanced Computing Center (TACC) and now AI & HPC technology strategist at Dell, among other career stops – is an HPC enthusiast with an outsized personality, a natural-born tech evangelist. Talk with him for 10 minutes and you want to go build a new high-speed interconnect that puts a dent in the universe. In this interview, he talks about his role at Dell aligning customers’ workload objectives with technology solutions, along with his background in HPC going back to the early 1990s.
Boisseau, who helps lead the Dell Technologies HPC Community, offers an observation about the common phrase, “the convergence of AI and HPC,” which is his increasing focus.
“I’m really interested in … the parts of AI that leverage HPC platforms,” he said. “We always talk about the convergence of AI and HPC … but it’s not quite correct. AI is really a set of workloads, and HPC is really a systems and architecture approach to enabling performance. And it is true that some of those AI workloads need HPC underlying it in order to get the required performance, such as deep learning training, of course…. Now I am beginning to branch a little bit on the AI side and understand more about what our customers need from the trained models on the inferencing side. Which is often not an HPC task for the application of those models, but it’s just as important. It’s the inferencing where the decision actually gets made that you train the model to make. So I’m doing a little bit of everything, in the AI space, but a lot of it is still focused on my historical background in the HPC space.”
