Expert Interview: Sudip Dosanjh

William Plasencia:  Hello there, this is Will Plasencia from Berkeley Lab Strategic Communications team. I’m talking today with Dr. Sudip Dosanjh, the director of the National Energy Research Scientific Computing Center, or NERSC, here at Lawrence Berkeley National Laboratory. We’re going to spend some time talking about NERSC and the future of computing. Dr. Dosanjh, how is NERSC shaping the future of scientific research? 

Sudip Dosanjh: NERSC has over 11,000 users running a really broad set of problems ranging from energy production to high-energy physics, material science, chemistry, and nuclear physics. So, they solve a very broad set of problems on our system. And as far as I know, we have the most users of any supercomputing center in the world. Our mission has expanded quite a bit over the last 10 years. Our users used to be focused on modeling and simulation, and what’s happened is that all the experimental facilities that the Department of Energy’s Office of Science funds are creating these large quantities of data and they need to be able to analyze that data. More data now is coming into NERSC than leaving NERSC. So, that’s a paradigm shift for a super computing center. When you move the data to a supercomputer, you need to look at different algorithms to be able to analyze it. The unique role that NERSC has is not just AI, it’s AI as a couple to data analysis coupled to simulation modeling. The users really want to run these very complex workflows, and that’s really what we’re targeting with our next generation system, which is NERSC 10, which will be fielded in two years.

Plasencia: How is the integration of high-performance computing and AI accelerating the pace of scientific discovery at NERSC?

Dosanjh: All the scientists have data, and they want to get more information out of their data. There’s a potential for AI to really lead to new scientific insights that wouldn’t have been possible otherwise. So, we have lots of users doing different types of energy research related to production, the development of catalysts. So, there’s really a very broad set of problems that people are solving. And all the experiments that people are doing, they’re generating more and more data, and there’s really a need for AI algorithms to be able to tackle that data. People at NERSC have expertise in things like computational science, data movement, algorithms, and so it’s kind of an ideal coupling. And we, of course, collaborate very closely with different companies to make sure our users can use the new systems that we’re deploying. 

We think quantum computing is really very promising for the future and will be a critical part of what NERSC does in the 2030s. We are working with a couple of different companies, QuEra and IBM right now, but we’re also looking at other potential partnerships. I expect to see a quantum system at NERSC that we would field in partnership with industry. NERSC is not developing quantum computers, but what we have is lots of users with very challenging problems, and it turns out about half the problems that people solve at NERSC are somehow related to quantum mechanics and quantum algorithms that have been proposed. So we think that there is a potential for our workload to adapt well to quantum computing in the future.

Plasencia: How could innovation in microelectronics make NERSC and other HPC centers more efficient and effective? 

Dosanjh: There’s a critical need for microelectronics research, both in terms of current microelectronics technology, but also new things like quantum computing. We’re kind of at the end of Moore’s Law. So it’s harder and harder to provide our users more computing capability within a given power envelope. We really need a strong push in terms of energy-efficient computing so that we can provide more capability to our users than they currently have. The countries that are able to solve this problem are going to have leadership in terms of science, national security and economic dominance. That’s really critically what’s needed. 

But then there are also things like quantum computing, neuromorphic computing – new paradigms for computing that could potentially provide a lot more computing power. Neuromorphic computing is an approach to computer design that mimics the structure and function of the human brain. The U.S. has always been a leader in terms of high-performance computing and its application to both national security and science and economic development. I think it’s going to take work for the U.S. to continue its leadership, in terms of investments, in terms of technology, in terms of applications, algorithms, that’s what the [national] labs are really great at, is team science, partnering with computer companies, with scientists, people who know computational science, mathematical algorithms, bringing everything together to be able to solve these very challenging problems for the future.

Plasencia: How would you describe the work that NERSC does that helps everyday people here in the United States? 

Dosanjh: There’s potential to design new materials through chemistry and material science for specific applications that can have a big impact on things like energy production; can have a big impact on manufacturing. You need material with certain types of properties, and there’s a potential to be actually able to design that material on a computer. Previously, people did trial and error, but there’s a chance to actually use AI to design materials for manufacturing, for national defense, for energy production with the properties that you need, and then you go manufacture it in the lab.

Plasencia: Where do you see the future of energy research going and the role that NERSC plays in that research?

Dosanjh: I think that right now, the biggest need is probably a concentrated effort in terms of AI. You know, there are lots of countries investing in AI, and there’s a huge potential for it to really revolutionize the way science is done within the Department of Energy that can have a big impact on national security. We certainly want to avoid technological surprises where some country leaps ahead of us in critical science and national security areas. In order to really make progress in AI, you need a big investment, not just in computing, but in people, people who understand science, people who understand mathematics and algorithms, and so you need team science, people with very broad backgrounds who can work together to push the frontiers of AI. I think we have an amazing set of people. We have scientists, we have mathematicians, we have people who know computational science and computer science. We have partnerships with industry. I think we have the potential to bring it all together and really take the next step in terms of AI for science; really enabling new modes of discovery that weren’t possible before.

Plasencia: Dr. Dosanjh, I appreciate the insights you’ve shared. To learn more about Berkeley Lab’s computing research visit nersc.lbl.gov. This is Will Plasencia from the Strategic Communications team at Berkeley Lab signing off.