The audacity incubator: 20 years of YFA | Ep 98

Intro Voices

Coming to DARPA is like grabbing the nose cone of a rocket and holding on for dear life.

DARPA is a place where if you don't invent the internet, you only get a B.

A DARPA program manager quite literally invents tomorrow.

Coming to work every day and being humbled by that.

DARPA is not one person or one place. It's a collection of people that are excited about moving technology forward.

Bart Russell
Programs usually do not last this long at DARPA. Why has YFA lasted this long? And it's because it's a continual, ongoing experiment of what is the best way to get the freshest, newest thinkers into our ecosystem as fast as possible.

Evelyn Wang
YFA was an example of disruption with which also resonates very much with the ethos of ARPAs, with disruption and thinking about how that is a good thing often to leapfrog and enable impacts in not on conventional in non-conventional ways.

Tim Haynes
Hello, and welcome to Voices from DARPA. I'm your host, Tim Haynes.

At DARPA, programs usually start and end on an accelerated cycle: nine months, two years, maybe even five-plus years for a complex platform program. And then they transition to another agency, military service, or an industry partner. Or we discover the limit of what's possible today, share what we've learned, and move on.

But for the past 20 years, one flagship initiative within DARPA has defied that mold. It's called the Young Faculty Award, or YFA.

Bart Russell
I'm Bart Russell, I'm the director of the Defense Sciences Office, which is the office dedicated to preventing and imposing scientific surprise.

So DSO runs the YFA program on behalf of the agency. We are the one to release the solicitation on an annual basis and run the contracts. We get submissions and topics submitted from the rest of the offices in DARPA, so we become the conduit by which the YFA reaches the rest of the agency.

Tim Haynes
Since its launch in the mid-2000s, YFA has awarded funding to more than 500 early-career researchers across the United States. And it isn't an ordinary grant. It doesn't ask individuals to write safe proposals that guarantee incremental success.

Instead, it invites academic researchers to take on bold, high-risk concepts before they even have tenure, pairing them directly with DARPA program managers, or PMs, as mentors.

Bart Russell
The magic in the program is the mentorship. You can get funding a lot of places, but it's not often that you have a mentorship program the way YFA works.

It's a little bit of matchmaking. There has to be an alignment of stars between the topics that the PMs are interested in in a given year, and the type of capabilities and the type of science that a proposer is interested in at the time.

Tim Haynes

In this episode, we're looking back at 20 years of Young Faculty Awards and looking forward to how the program continues to evolve to grow the pipeline of researchers solving our nation's most pressing national security issues.

In addition to talking with Bart, we'll hear from three YFA alumni who went on to leadership roles across government, academia, and industry.

Bart Russell
The YFA program is primarily focused on boosting the careers of early-career scientists who are thinking in exceptionally fresh ways that align with our PM needs and giving them a fast track into the DARPA ecosystem.

These aren't grants. I call grants like fire-and-forget money, and that's not what DARPA does. These are milestone-based awards, and so the academic has to deliver against those milestones in order to get paid. And that in itself is training, because a lot of academics are used to grants, and this kind of contract relationship is really different, but it's how DARPA has to work in order to deliver our products and in order to fulfill our mission.

Working with the PM, some of the PMs will invite that YFA awardee to come join their PI meetings for their other programs, much larger programs. And then through those avenues, the YFA awardee can get exposed to other performers they might want to partner with, to broader problem sets. They might be invited to major demonstration and training events where they're at a military base and their helicopters are flying over their heads, and that's not the typical kind of environment you see in a lab. 

So when we do this matchmaking, they get exposure to other sides of industry. They can then maybe get inspired to do a startup. They might work directly with one of those companies, and that's the ecosystem that we see daily that a lot of our YFA awardees don't see until they get embedded into our work. Tim Haynes

To see what those ecosystem connections look like in practice, it helps to see where YFA awardees land 10, 15, even 20 years later. In 2007, in the second year of YFA, one of the recipients was a young mechanical engineer who had just joined the faculty at MIT.

Evelyn Wang
My name is Evelyn Wang. I'm currently serving as a Vice President for Energy and Climate at MIT and also am a professor of mechanical engineering.

I first heard about DARPA through my advisor when I was in graduate school. Actually, my advisor was Tom Kenny, and I was co-advised by Ken Goodson. Tom, in fact, was a program manager at DARPA when the YFA launched. So that's how I got to know about the YFA program, but really started from the origins of when I was working on microchannel cooling, thermal management for electronics for a DARPA-funded program more than 20 years ago.

Tim Haynes
For Dr. Wang, the award was pivotal. Beyond just financial support, the program gave her a strategic compass.

Evelyn Wang
The YFA was the first federal grant that I got as a starting faculty member. And so that really was critical to the rest of my career so far, because as you likely know, it's so hard to get that first grant. And getting that, I think built a lot of confidence, but also curiosity in the opportunity space in terms of the kinds of research problems we're looking at.

And the thing about DARPA is that it's very mission-driven. That was one thing that really resonated with me in thinking about the problems that we want to solve and why it matters. And I think that really helped me as I pursued my academic career and pursuing faculty research in areas that I think would make an impact.

And I still remember my mentor was Dennis Polla, and he gave me great advice and really helped me navigate what are real applications based on the access he had to the various challenges in the DOD, and seeing where the work that I was doing that was still early stage, how it could be tailored to some of these applications if I was successful in developing the technologies.

And I think that was hugely valuable, as well as the cohort that evolved from having a lot of different expertise of researchers getting these awards, and being able to meet them and understand the different kind of opportunities and challenges they also have, and thinking about, well, how does my work and my solution fit within the context of the broader ecosystem, not just doing academic research, but being able to serve in a larger capacity to help make impacts for the nation.

Tim Haynes
In 2022, her journey came full circle when the U.S. Senate confirmed Dr. Wang to lead the Advanced Research Projects Agency for Energy, or ARPA-E.

Evelyn Wang
I had the privilege of leading an agency which was modeled after DARPA, called ARPA-E, focused on disruptive energy technologies. And I learned a lot from the DARPA funding that I received starting from the YFA, and how we can take advantage of a model which is very much looking at high-risk, high-reward, impactful technologies that could really transform a particular area of need.

And that's what I took on as I led this agency to be able to think about blue sky opportunities by which we could really make an impact, and be able to transform the future of energy, in this case, and really inspired by all the things I've learned along the way through DARPA.

Tim Haynes
Today, Dr. Wang brings that same high-impact mindset to her role as Vice President for Energy and Climate at MIT, uniting academic researchers with government agencies, industry leaders, and policymakers.

Evelyn Wang
To work on problems that matter is so key to how we can make a difference as academics and whatever other career paths that we take along the way.

It's about being continuing exposed to new problems, being able to actually identify what are the key bottlenecks to problems so that as a researcher, we can actually address those bottlenecks, rather than making up problems that actually aren't real and that actually don't matter.

And engaging beyond what we know in the academic world and have exposure to real systems, real problems in industry, in the government, in the armed forces, all of this matters to help shape the directions that really move the needle to help build a better future for people and the planet.

Tim Haynes
While Dr. Evelyn Wang's story demonstrates how the YFA can help launch a two-decade trajectory to national leadership, the program's value to newer faculty members is just as pronounced.

Suo Yang
I'm Dr. Suo Yang, and I'm currently an associate professor of mechanical engineering at the University of Minnesota, where my lab's name is Computational Reactive Flow and Energy Lab.

We focus on multiphysics and also AI-assisted modeling for turbulent reacting and multiphase flow, which include a lot of topics like combustion, non-equilibrium plasma, particulate and gas-liquid flow, and hypersonic.

Tim Haynes
When Dr. Suo Yang looked at funding opportunities across the government in 2020, he noticed that most young investigator programs featured broad, generic topics, whereas established labs often held an invisible advantage. But DARPA's YFA was different. 

Suo Yang
Other programs they are more like open topic. DARPA YFA, at least for my years, were very specific, targeted topics. And when you read it, these topics are specially designed and sometimes you feel it's so uncommon and unconventional, but that's have one consequence: almost no one can claim they have a significant track records on the topic because it's so uncommon. 

That's effectively reset everyone back to the same starting line, and so it's incredible for early career researchers like me, because I won't have someone say, "I did that for decades and no one can beat me." Right? So that's initially why I decided to apply for YFA. 

Tim Haynes
One of the problems DARPA posted was an engineering challenge that had sat virtually untouched for nearly 50 years: finding a new theory to predict and model sonic booms generated by flying at hypersonic speeds of Mach 5 and beyond.

Dr. Yang's DARPA mentors didn't just let him simulate the math on a computer. His first PM mentor, Dr. Andrew Baker from the Tactical Technology Office, pushed Dr. Yang to test his models against real-world physics, sending the academic team to Buffalo, New York, to conduct high-speed wind tunnel testing at CUBRC. There they produced some of the first hypersonic, sonic boom measurement data in decades.

After two successful years of mentored research, then came the next step built into the YFA structure: the Director's Fellowship. Awardees who meet their aggressive milestones can compete for an optional third year of funding. In Dr. Russell's words: 
Bart Russell
The Director's Award is extremely useful when there is an exceptional opportunity to align what really is sort of pilot or seedling work, and either get it injected into a larger program or get it more runway to test whether a small idea can take the next step into the real world.

It's as much money as the base award, which is incredible, and it shows a commitment that, no kidding, we really like this and we want to see where it can go.

Suo Yang
It's like a catalyst for me. YFA have this special structure of 2+1. You got a two-year award, then you compete for a third year, which is called the DARPA Director's Fellowship Award.

Fortunately, I successfully got the Director Fellowship Award, but it's far more than the project itself. The DARPA YFA was my very first DOD grant, and it helped me gain a lot better understanding of the DOD research ecosystem. And that unlocked a series of critical support I got later, which include in a big project from Army Research Lab on the UAV propulsion system control, and also I got two young investigator awards: one from Office of Naval Research, one from Air Force Office of Scientific Research.

Tim Haynes
That cascade of funding didn't just earn Dr. Yang tenure, it reshaped his trajectory as a defense science leader. Just this past year, Dr. Yang was named a Chief of Naval Research Fellow with the Office of Naval Research, helping advise the Navy on future propulsion and power research.

When asked what he tells other young faculty members starting their careers, his advice reflects the DARPA ethos:

Suo Yang
Don't try to play things safe. If a classical thinking or ideas limits what aerospace or propulsion or whatever system can achieve, go back to first principles and think about underlying assumption and see whether you can push beyond that. Second thing is always ask yourself, so what? DARPA cares about fundamental science, but they care as much about the capability. Science is good, but why it matters, what capability can bring for both United States and the field, you need to think about that.

Third, if you have something no other funding agency are interested in funding it, because either it's too advanced or too early phase, and you probably want to consider DARPA. This is a right place for out of box thinking and for high-risk research.

And finally, don't be afraid of risk. I would suggest embrace the risk of bold ideas.

Tim Haynes
The willingness to push beyond conventional thinking is perhaps nowhere more visible today than in the field of machine learning, which is closely tied to related terms like artificial intelligence and superintelligence.

Zico Kolter
My name is Zico Kolter. I am a faculty member and department head of Machine Learning at Carnegie Mellon University, and I work pretty broadly in AI systems. These days doing a lot of work on AI safety, AI security.

Tim Haynes
When Dr. Kolter became a YFA fellow in 2017, deep learning was just beginning its explosive commercial rise. Most researchers were focused on standard architectures. Dr. Kolter used his YFA award to ask a contrarian question: What if we could place entire optimization algorithms directly inside of neural network layers? 

Zico Kolter
So my work under the YFA was on optimization in the loop of deep learning models. What our work did was ask the question of could we have a more complex operation, like actually solving an optimization problem, maximizing or minimizing some objective inside a network, and make that process part of the network itself? 

This sort of mindset of thinking about optimization processes as units within deep networks, as sort of a core capability that you want to look at within deep networks, has been really, really influential in my work and in my group's work. 

At the dawn of the LLM era, we sort of leveraged that same kind of work to build techniques that could do things like jailbreak modern large language models, so essentially get them to bypass whatever safety training they had and access some of the capabilities that developers did not want to expose. 

That particular direction has led to a whole lot of work that I've then started doing broadly in AI safety, both in an academic setting and well beyond. Some of my colleagues on that work and I started a company, Gray Swan, which is now a 60-person or so startup working in AI security, being heavily used by a lot of the frontier labs, a lot of enterprises. 

Somewhat as a direct outgrowth of this was invited to join the OpenAI board, where I chair their Safety and Security Committee, which oversees the governance of a lot of their model release process and sort of the safety aspects of that. 

There's not a direct chain between any of these things, of course, but I think a lot of the work here kind of led to our work in robustness, which then essentially motivates what I do now, most of my work in AI safety and security.

Tim Haynes
From a YFA project in 2017, now Kolter's in the boardroom helping oversee safety for some of the most powerful machine learning systems on Earth.

When Dr. Kolter looks back at the role DARPA played in supporting that journey, he points to something unique about the agency's culture: its tolerance for research that doesn't follow the safe, predictable path.

Zico Kolter
The nature of science is a process of starts and stops, of attempts to try to get something to work, failures, successes, building on those failures and successes. That is the nature of scientific progress.

And for a very, very long time, in a very fundamental way, DARPA was one of the few agencies that was willing to take those risks and be okay with failure, which sounds really funny because, you know, we're highlighting things that led into that have been arguably somewhat successful, but there have been many more failures of research projects. It is much better to attempt something audacious and fail at it than it is to make sure that you're going to deliver on the exact thing and number of things you promise.

It really is about the fact that as long as you are in close contact, explaining what's happening, explaining sort of the developments you're making, the people there are hugely supportive, have been really instrumental to helping shape the entire nature of the field, and it's been an amazing place.

Tim Haynes
Always attempt the audacious. For two decades, that philosophy has allowed the Young Faculty Award to serve as an incubator, not just for ideas, but for people.

And as Dr. Russell points out, when you invest in an early-career academic, you aren't just funding an individual researcher: you're planting the seeds for an entire academic family tree.

Bart Russell
The generational impact story, that has been a surprise. You know, we originally thought about this as jump-starting the career of a single PI, but more and more what we're seeing, and it makes sense in retrospect, that PIs never work alone. They work with their graduate students, and those graduate students are getting exposure and, you know, getting the bug of working with DARPA, because it is an extremely fun place to work. There's a lot of fun problems to work on. And in so doing, it's inspiring those graduate students to want to go after YFA.

I just heard from one of the other office leaders yesterday that the YFA that they had mentored when they were a PM not only used the work that we had supported as the basis of their tenure argument at a pretty prestigious, leading university, but also that this past year his graduate student got a YFA award.

It's a wonderful example of having positive impact on that academic's career, but then also that it's not just that person: it goes down throughout their academic family tree.

Tim Haynes
To build on that momentum, DARPA has continued to adapt the program. Last year, DARPA introduced a targeted option to the YFA solicitation: dedicated supplemental funding specifically for awardees to support a U.S. citizen graduate student in their labs.

Bart Russell
This isn't an either/or. We're not asking you to not choose another superstar in your lab to work on the project, but rather the incentive to bring somebody who is immediately clearable into the lab to learn how the science works, who could then carry that forward into the defense ecosystem more naturally.

The experiment is that by putting this extra money directly in the lab of a PI who's already demonstrated themselves to be extremely capable and forward-thinking, and have an extra graduate student studying under that person on a topic that we already know we care about, bolsters that pipeline in a very specific, targeted, and efficient way.

These are going to be pilot data to ask a question of whether a program like this should be broader. If we can get some really compelling data off of this experiment, there's no reason why I couldn't go to other agencies and show them the model and say, "Hey, maybe you want to do this, too."

Tim Haynes
The Young Faculty Award remains an ongoing experiment, and one that continues to evolve to meet the needs of the moment and have an outsized impact for both the public and private sectors.

Bart Russel
The past 20 years is a grand success in getting new thinkers into the ecosystem.

So I like to say that when you're working in science and on problems like that, it's not that we have the farthest-off horizon, but rather we have the biggest lever arm. And with the biggest lever arm, we can have not just impact for one kind of self-driving car, but rather create an entire new ecosystem in industry that just doesn't exist right now. That's the kind of power we're offering. And if you get a YFA, it shows that your idea could just be that kernel, that nugget that leads to that kind of revolution.

Tim Haynes
Whether it's cooling the next generation of high-power microelectronics, rewriting the physics of hypersonic flight, or safeguarding the frontiers of superintelligence, the past 20 years of the Young Faculty Award show what is possible when the nation wagers on the curiosity and courage of early-career thinkers.

If you're an early-career researcher with a bold, unconventional idea to solve a national security challenge, keep an eye out every fall for the next year's YFA topics to look for a project and a mentor aligned with your interests.

To learn more about the Young Faculty Award and 20 years of the YFA fellows and their work, check out the show notes or visit darpa.mil.

Thank you for listening to this episode of Voices from DARPA.