Margaret Martonosi, a pioneer in computer architecture research, has recently returned to campus after spending the last four years leading the Directorate for Computer and Information Science and Engineering, one of eight divisions of the National Science Foundation.
Martonosi, the Hugh Trumbull Adams ’35 Professor of Computer Science, has spent five of the last 10 years working in the federal government. In 2015 and 2016 she served as a technical adviser at the State Department as part of the Jefferson Science Fellowship program, which brings senior science and technology experts into international policy discussions.
In the interview below, Martonosi explains what her service at the NSF entailed and reflects on some of her experiences in Washington. While Princeton can have amazing influence, Martonosi said, during her time as a leader at the NSF she learned to think even bigger, creating policies and programs that impact scientific inquiry and improve academic research more broadly at the national scale.
You’ve spent half of the last decade working in the federal government. Why did you choose to pursue public service?
Like a lot of computer scientists, I appreciate the societal impact our ideas and work can have. If you look at the broader throughline of my work, I have always been interested in societal impact and working at the national and international levels.
In the years leading up to my fellowship at the State Department, I had done some work on low-cost internet connectivity and bringing internet technology to developing regions, for example. I did work in Kenya on wildlife tracking. I worked on deploying a technology for low-cost internet access in Nicaragua. I was doing international work in my own way, and that fed into the work I did at State in international policy.
With NSF the connection was a little different. I had been on the NSF’s CISE (Directorate for Computer and Information Science and Engineering) advisory committee, and I knew how much the NSF means to our research community. NSF provides 80% of the federal funding that goes to academic computer science. Basically, you don’t have an academic career in computer science in the U.S. unless you get NSF funding. For me, working at NSF was a chance to say thank you for how much they had supported my career at key moments.
I also wanted to make it possible for more people to have NSF support. Working to broaden access was key. It shouldn’t just be for lucky people or people at the top schools.
NSF is also a really great place to work. It’s consistently listed as one of the best places to work in the federal government. And while I enjoyed my time at the State Department, there the scientists are outliers. At the NSF, the scientists fill the majority of the positions at all levels and play key roles in all aspects of the agency.
What did you do as the head of the CISE Directorate at the NSF?
My role was to express technical strategies regarding NSF’s support for the CISE research community, and to work through federal budget processes to help make those funding visions implementable. This included both internal-to-NSF and inter-agency technical discussions as well as hearing from the community about what was needed and what their ideas were. Also looking broadly at the field and the NSF portfolio and trying to understand where we need to go as a field and as a nation.
For example, while the NSF was already making a lot of investments in AI before I arrived, during my time there the National AI Research Institutes program made its first awards, and over time went from zero to 25 of these large-scale research investments. Each AI Institute is a $20 million investment over five years, with about a dozen universities involved. CISE leads that program, but several other directorates at NSF contribute. For example, an AI institute for weather and climate got co-funding from the geosciences directorate; an AI institute on materials science got co-funding from the math and physical sciences directorate, and so forth.
Another big piece of the job is formulating future-looking ideas for what the NSF could do with appropriated funding. If NSF gets direction from Congress or the White House to, say, spend more on environment issues and research, we would be ready with ideas for how to do that. All federal agencies receive a budget and direction, a process which involves both the White House and Congress. During that budgeting process, leaders of directorates try to square the things that we know the country needs with the constraints that are placed on us by the overall budget and the directions that are given.
Part of this involved illustrating the opportunity costs that come with not reaching certain levels of funding. NSF has so many amazing stories of impact and helping people and we need to get better at making sure the average American taxpayer is aware of those stories. I can show you the annual grant report provided to NSF where Stanford professor Hector Garcia-Molina wrote, “Oh, by the way, while being funded by this grant, we did a little startup. It’s called Google.”
If you’ve used Duolingo, that has NSF funding in its roots. And some other AI key contributors like Databricks, too. These are all companies that have come out of NSF CISE investments. Being able to tell those stories effectively and show the impact of funding this research is very important.
The role is also about running a large organization with a $1-billion budget and about 150 staff members. It’s on par with a dean-level position. It was a great experience. There’s an amazing mix of rotators, i.e., people like me who come in from the research community, working alongside long-term federal employees who are extraordinarily dedicated to their jobs.
In addition to the AI Institutes, are there one or two other projects you’re particularly proud of?
Yes, here are two other projects I’m particularly proud of.
I started at my role at the NSF in February 2020, five weeks before the COVID lockdowns. The spring semester is huge for faculty interviews and hiring in computer science, and the pandemic made everything very uncertain. Some schools continued their hiring process on Zoom, others canceled their searches. Our worry at NSF was that, after funding Ph.D. students for four or five years, these folks would be forced by the uncertainty to take other jobs and possibly leave academia and research forever.
On essentially zero notice we created a postdoc program called CIFellows in 2020. A program like it had been launched in response to the 2008 recession, so we had some idea of how to do it, but we had to get it up and running very fast. With great partnership from the Computing Research Association, the fellowship program was up and running by May 2020, and we had postdocs in funded fellowships by September.
One thing I’m really proud of is how they matched the fellows to host institutions. They used what’s called a “Max-2” approach. This ensures that no more than two fellows should have gotten their Ph.D. from the same school, and no more than two hosts should be from the same school. This meant that it was a really rich mix of people and schools benefiting from the program, and it covered the whole country.
The other project I would highlight is ongoing — the CSGrad4US Graduate Fellowship. A lot of people have heard about the NSF graduate fellowships, one of the NSF’s oldest and most competitive programs. CSGrad4US is similar in terms of funding amount, but it specifically aims to increase the number of U.S. citizens and permanent residents studying computer science at the doctoral level.
The number of U.S. citizens and permanent residents going to graduate school in computer science has been dropping steadily for well over 10 years, a trajectory that’s worrisome to many people. Roughly two-thirds of doctoral students in computer science are now international.
While there is great value in welcoming doctoral students from all over the world, it’s also important that folks who were educated in the United States understand the opportunities that research careers offer and get appropriate mentoring about how to pursue them. For example, at many schools in the U.S., the computer science programs are so big, with so many people flowing through, that promising students aren’t always getting the attention they need. We need to make sure we are pulling folks aside in the fall of their senior year and saying, “Hey, now is the time to apply for graduate school and fellowships.”
Also, student loans are a big factor. Students here in the U.S. often graduate from their undergraduate programs with huge student loans. And many of them, on the day after graduation, can go to industry and be earning a salary that’s the highest their family has ever seen. There’s a whole bunch of financial reasons why working may be the choice they need to make instead of going to graduate school right away.
The goal of the CSGrad4US program is to help people working in industry find a pathway back into graduate school. The program gives fellows one year of mentoring to learn about doctoral degrees and their application processes, and then three years of fellowship support at the same level as the NSF graduate fellowship program.
Did anything surprise you about working in the federal government?
Working at the NSF brought home to me just how much the different federal agencies work together.
I chaired the NITRD (Networking and Information Technology Research and Development) subcommittee, a broad inter-agency subcommittee that thinks about all aspects of computing research and development. It includes dozens and dozens of participating federal agencies. The Department of Energy, NIST (the National Institute of Standards and Technology), the Department of Defense and more. At the meetings, you’d also have someone from the National Archives on the call, someone from the Department of Commerce, someone from DARPA (the Defense Advanced Research Projects Agency).
Conversations like these span all the federal agencies, with different people bringing in their skills and their awareness of different opportunities and challenges. I think that was huge and important.
For example, the CHIPS and Science Act was passed while I was at NSF — $50 billion appropriated to re-onshore the semiconductor industry. I was on a conference panel about a year ago with some other federal agency representatives to talk about the legislation. Beforehand, a university colleague in the audience asked if I had met the other panelists. I knew them well! I was on Zoom calls with them twice a week! We wrote joint reports about semiconductor research and development. These reports are public, but for some reason it isn’t obvious to people how much inter-agency collaboration is happening.
What lessons have you brought back to Princeton from Washington?
My time at the NSF and at the State Department was really valuable, and I hope Princeton sees that and will be even more ready to help the next folks go do rotations like this. Given the motto of our University — in the nation’s service and the service of humanity — there needs to be support for things like this. I have spent more time in Washington than most academics. I’ve found it very rewarding. I’m one of the few people who can say they’ve served the last three presidents.
Since coming back, I’ve had a lot of fun having small group or one-on-one conversations with faculty who are just starting to write NSF proposals, to give them some sense of how to get started. I can also help explain the alphabet soup of acronyms and programs — it’s not easy to navigate.
Also, at a higher level, I can help people be more aware of the budget situation and how it affects research funding. For example, some faculty members who I consider to be extremely tuned into what’s happening in Washington hadn’t realized that NSF got a budget cut this year. If we don’t speak to the importance of NSF’s budget, then who will? And if we aren’t watching it, then we can’t speak to it. But it can be hard to devote the time to tracking this news.
The part that has really changed me is thinking beyond Princeton. Princeton can do things with amazing impact, but often not at a large numeric scale. How can we help science as a whole and hundreds of universities at once? How can partnerships with other universities help us do things at scale? I think it’s important for us to have more conversations about this.
Martonosi’s research focus is computer architecture for classical and quantum computing. She is a fellow of the Association for Computing Machinery (ACM) and the Institute of Electrical and Electronics Engineers (IEEE), as well as a member of the American Academy of Arts and Sciences and the National Academy of Engineering. She was the 2023 recipient of the ACM Frances E. Allen Award for Outstanding Mentoring, and the 2021 recipient of the ACM/IEEE-CS Eckert-Mauchly Award, the highest honor in computer architecture.