With most systems and architecture conferences taking the online route, we figured it’s a great time to get to know a few people in the systems/architecture research community. People of Systems & Architecture is a series of interviews conducted since last year, and continues in the same vein as the People of PL, People of POPL, and the People of Language Design and Implementation interviews done by John Wickerson, Jean Yang, Brandon Lucia, and Minjia Zhang.
A: Tell me a little bit about your career journey.
K: We were undergrads in Brazil at the University of São Paulo. Towards the end of our undergrad, we decided we wanted to travel to the USA to see what was around. We visited the USA along with six other colleagues. We went to different companies and universities just to see what we wanted to do after we graduate. During this process, we met the person who ended up being our manager at IBM research, José Moreira. He came to Brazil, saw our graduation project, decided to make offers to both of us, and brought us to IBM research.
That ended up being a really eye-opening experience. It was supposed to be a three-month internship, we were able to extend it to five, but suddenly 13 months just went by. This opportunity really gave us an understanding of what research is about, and how to learn and investigate different things. We are both lifelong learners; we always love to learn about new things. This research experience motivated us to start our PhDs at the University of Illinois.
L: We have always worked really well together. We have very complementary skills and complementary ways of thinking too. In grad school, we did some work together and then each focused on our own thesis. When we graduated, I wasn’t sure if I wanted to be in academia, so I applied to both academia and industrial research. Karin was sure about wanting an industrial research position. We are so glad we managed to arrive at our current positions, where I get to be at U. Washington and Karin at Microsoft Research, and we still collaborate a lot.
A: How did the DNA storage project come about?
L: We’ve always had a taste for unusual topics. We specifically decided to focus on biology because nature is such a great source of fantastic mechanisms. We had met Georg Seelig, from whom we learned a lot about DNA nanotechnology, and played with DNA computing a bit before our sabbatical in Cambridge. During our sabbatical, we saw Nick Goldman talk about their DNA efforts and the need for better storage. We thought “Oh, that’s really cool, but how do we think about this from a systems perspective?” That’s when the whole project began. The next thing we knew, we were back from sabbatical, and had started doing experiments. Microsoft also started becoming interested in this work. The project moved really fast after that. We thought about the problem from a systems and architecture perspective, while other groups were focused on individual aspects. I think that was the unique thing about what we did.
K: Rather than focusing on a single component, we really wanted to understand how the whole thing was going to work. I also want to mention the great mentors who spent time educating us. That made a really big difference. For example, we learned enough about DNA computing from Georg Seelig to be in a position to appreciate the talk at Cambridge when we saw it.
It’s great to meet people, work with them, and then continue fruitful collaborations.
A: It’s always rather challenging to begin working in an area that others haven’t really worked on in the past. What are some challenges you faced when you wanted to combine the biological aspects of DNA with systems research?
L: The biggest thing is a communication challenge. We needed to make it clear that we weren’t doing it just because it is cool. There’s really a big potential here. The most important thing is to communicate this high-level message and avoid the unnecessary details that can obfuscate this main message. It’s important to find the right language, the right analogies, and the right place for this potential opportunity. For example, thinking carefully about when it makes sense to use molecular systems vs. just using them for everything.
K: I agree.
When you’re working at the intersection of different fields, it’s not just communicating to the community you’re in (like computer architecture for us), but trying to communicate with the intersecting communities as well.
Establishing that common vocabulary was one of our biggest challenges. With our diverse team, we found that even the mentality of how you set up an experiment was different in computer architecture vs. biology. So, it was important to harmonize all of that, and understand where the different people were coming from. Once everyone got on the same page and started speaking the same language, we found that we progressed a lot faster.
A: What are your thoughts after receiving the Maurice Wilkes award?
L: We were happy obviously, but we weren’t expecting it. We love what we do and we are incredibly passionate about it. But, there are so many well-deserving people doing so much great work in the community. We can think of plenty of other people who deserve the award as well.
K: This has really been a team effort, so I consider the award as a recognition of the great work done by the entire team, not just the two of us. That’s how I like to think of it.
A: One thing that is very unique to you is that you always tend to work on topics that go on to become the next big thing. For example, we notice this trend with the race detection, DNA storage, and TVM projects. Even your students seem to follow a similar trend (e.g., Brandon Lucia with the intermittent computing in outer space project). I wonder if you consciously try to figure out the next big bet? How do you envision your research vision and pick your research problems?
L: Oh wow, thank you. That’s really flattering.
I tend to think very consciously about what the big opportunities are by looking at trends early.
I think making early bets is possible by always being attuned to technology trends, application trends, and identifying opportunities that intersect various other fields with computer science.
For example, Karin and I haven’t really done traditional architecture in a while. This is also why ASPLOS is our favorite conference. It allows us to focus on intersectional areas.
With the DNA storage project, the trend was that the progress in reading/writing DNA is just super fast. At the same time, there are indications that storage devices are losing steam in terms of scaling. That was the specific trend that helped guide our research direction.
I’m a firm believer that, especially in academia, we have got to place bets. It is super important to focus on the more established problems as well. But what keeps me excited is exploring really new things. From my personal viewpoint, I think that’s what academia is really about. But, I should also add that a lot of these efforts also fail. In hindsight, it’s easy to say that we were able to make progress on several new things, but people generally don’t know about the multiple failed bets.
K: Yeah. It’s a little bit of selection bias.
Both Luis and I have had the habit of reading broadly, not necessarily in just computing, but multiple other fields. We find such reading/discussions super fun. So, we’re motivated in a way by the fun, but that has also helped provide more context in other areas and opportunities for the intersection that Luis was talking about.
L: Also, never forget that research is a very social experience too.
You have to form a team with the right set of people. It’s very hard to do big things, especially in the intersection of multiple areas, in isolation. So, we collaborate very broadly. Brandon Lucia was the only student I’ve had a paper with where it was just me and one student. Almost all of my works have multiple collaborators, often from different fields. This applies to both senior and junior co-authors as well. Without these collaborations, it’s impossible to know everything and learn everything fast enough to go where you want to go.
A: Does being in Microsoft Research help you know about these real-world industry problems and stay ahead in terms of knowing about these trends?
K: Yeah, definitely. That’s one of the things that attracted me to industrial research: to understand real-world problems and figure out creative out-of-the-box ways to tackle them. The convenient thing about industrial research is that you know about the problems that industry has right now, but you also tend to know about what’s coming up. That’s why I was so sure I wanted to be in industrial research when I graduated.
A: What is your secret to being able to work at the intersection of so many different areas?
K: I think of this portfolio of research and projects people work on as sort of a “T-shaped” thing.
You want to go broad, but, you also want to go deep in at least one topic, which then becomes your main topic. When going broad, you sort of poke at different areas. Once you find an area in this breadth that you want to work on, then you have to find somebody who is deep in that area and has that overlap of the breadth that you have. That’s how our collaborations have come to work.
A: I also notice that a lot of research you do goes on to become big in the industry (e.g. TVM). Do you explicitly work to get your research inducted into products?
L: That’s a good question. I would say that it’s a mix of different things. With TVM, we talked to several potential industry collaborators. Then, Amazon joined as an early collaborator. These collaborations with industry certainly help guide the work. There was the trend component here as well. We found this gap between machine learning models and the hardware they run on. So, we were able to start from the beginning, which meant that there was a lot of transparency to the stuff that other people were doing.
But yeah, I’d say that just making sure you talk to industry often, and keeping track of the interesting trends, will help you end up with solutions that many people have a use for.
I don’t think we should only solve the needs of today. We started working with TVM before model compilation was important. We made a bet that it would soon be important to have an automated approach to compilers. In hindsight, TVM worked out and we were lucky, but we might not have been as lucky as well. There’s a quote I really like:
The best way to have good ideas that go on to work well, is to have a lot of ideas so that you find some that stick.
This means that you have to be willing to fail a lot in research.
K: It is really important to understand the context around the problem you’re solving. It’s important to have many ideas, yes, but weeding out the ones that are not going to pan out is just as important. Identifying trends such as growth in compute power, memory and storage capacity, network bandwidth, etc., will help with the weeding out process since these trends can help project how the problem and solution are going to evolve over time.
A: How did the OctoML startup come about? What advice do you have for how to manage academic/industrial research along with a startup?
L: For how the company came to be, TVM was adopted as an open-source project. There was a clear opportunity to get the team behind TVM and the momentum behind the project to convert it into a product. So, we had the right team, TVM was getting adopted widely, and the stars also just aligned. I became a full professor and am on sabbatical now. It was just too hot, and too good of an opportunity to just let it pass.
In terms of managing everything, I think it’s key to form a great team. I’m a firm believer in forming the best team I possibly can, to have people I can trust with job delegation, so that I can remove myself from a lot of stuff that’s in the critical path.
I think it’s a combination of finding the right way of configuring things, getting out of the way when needed, empowering people to do things, being really good at delegation, and being grateful and fair to the people you work with.
It is key to make sure that everyone in your team has everything they need to succeed.
A: What do you like the most about your job?
K: I like learning. I like the fact that my job requires me to constantly keep learning about new things. Once you’ve learned something, the obvious next step is to apply it somewhere to solve a problem or make someone’s life better.
L: When learning something new, there’s some uncertainty to things which I find quite fun. It’s an intellectual thrill to dream about outcomes when learning something new. The best part about learning something new is getting to meet new people. I just love meeting people, learning how people think, making new friends, and meeting people that think differently. Almost every time we started a new project, there were new people involved, and I had the opportunity of creating new relationships.
A: How do you think things would have turned out if both your roles/positions had gotten interchanged?
L: Oh, that’s a great question! It’s hard to say. I think we would still work together, there’s no question about that. We think of research as an adventure and we love working with each other. If the arrangement wouldn’t have worked out, we would have probably flipped our roles.
A: What are some strategies that have worked well for you in terms of working together as a team?
L: First of all, we both have similar tastes for the kinds of problems that we like to solve. We are also very well-aware of each other’s strengths and weaknesses. So, when working on something, our roles more-or-less are organically defined. It’s not that each of us always does the same thing, but we evolve our roles based on what each of us does best. Over the years, we have achieved the ability to complement what the other person brings to the table. The fact that we know each other really well, and have a shared life vision also certainly helps.
K: That said, I think a similar collaboration can also be established between people who don’t know each other that well. It goes back to the communication aspect I was talking about earlier. It is important to set expectations, be clear about different people’s roles, make sure that people complement each other, and have each person focus on one important area.
When establishing a new collaboration, the most critical thing is to establish a good communication channel.
A: When I was interning at Microsoft Research as a second year Ph.D. student, and I met with both of you, I still remember all the incredible advice you gave me on how to be a good researcher, pick a thesis topic, etc. I benefited a lot from those conversations, so I certainly want to touch upon that here. What advice do you have for junior researchers and students?
L: The first thing is to talk to people about your research. Just go talk to them. Don’t be afraid. Especially talk to people whose work you find interesting, irrespective of whether it relates to your work or not. Remember, you are a part of this systems community.
Being a part of a larger community is what makes research fun and interesting.
K: I just want to interject a bit here. Definitely talk to people, yes, but also definitely do not believe everything they say. Sometimes, you might get some conflicting advice that might not necessarily make sense. At some point, you’ll need to do some critical thinking of your own based on everything you heard and figure out the things you agree with. If you don’t agree with something, ask questions and clarify things. Then, put everything you’ve heard together in your own context and create a bigger world view for yourself. This could be in terms of your entire career or a project, etc.
L: Yes, absolutely. Also, I would again emphasize looking at and reading about technology trends; this is very very important. Do things that interest you because this will help things “click” in your head. When your brain is excited about something, that’s when you become most creative. For example, I like cooking and looking at art. When I do these things that I find interesting, I find that it puts me in the right frame of mind to come up with new and creative ideas.
There’s one more piece of advice that I’m not sure if everyone will agree with me on. I’m not sure if Karin will agree with me either. But, I honestly believe that sometimes, as academics, we are very cagey about talking about our research to others because we believe we might get scooped.
When I look back, I find that I have always benefitted from talking about my research with others.
Yes, I have been scooped a few times, but, on average, I think I have benefited more from getting early feedback on an idea. So, if you have an idea, I think it’s certainly much better to just go and talk to people and say, “Oh, what’d you think of this?” I generally talk to people about the idea, write my thoughts and share them, write workshop papers, write blog posts, get feedback, and close the refinement loop as early as possible before I put in a bunch of work. I think this process helps me hone down on the impactful ideas.
K: At the end of the day, as a PhD student, you just want one key contribution that’s your own, is significantly impactful, and is something that people remember you for.
This is much more important than writing a ton of papers.
A: @Karin, what is like to be a woman in the architecture and systems community? @Both, as a community, what can we do to improve inclusion, diversity, and equity?
K: Well, let’s just say that there aren’t that many of us in the field. I’d love to see more representation. In fact, I wanted to highlight that we donated the Maurice Wilkes award to “Black Girls Code”. I encourage everyone to think about how to bring more diversity to the field.
There are lots of research opportunities that are lost because we don’t have the right representation of researchers in the team.
The diversity of backgrounds we have as a team doesn’t just refer to technical diversity, but really all kinds of diversity that can help bring in different viewpoints/perspectives. Luis and I are putting a lot of our energy in solving this problem.
I think that we have a real problem with the pipeline. One initiative that I thought was pretty awesome was an initiative by Bobbie Mane, Lena Olson, and Newsha Ardalani, who created an undergrad workshop at ISCA. So, undergrads get to come to our conferences, and see what research in our area is all about. They get to talk to different people in the community and understand what “systems” is. There are also plenty of other opportunities to get the broadest, most diverse set of people excited about technology. Retention is another area where we could do a lot better: nurturing and checking in periodically with diverse people working in our area, to make sure their career is progressing as expected and they feel supported.
L: Yeah, I completely agree. I think we need to make a conscious effort to solve this in a distributed fashion, where everyone has to do their part by being welcoming and encouraging. This is the very least that we must all do. But, we should also be doing a lot more to consciously offer opportunities to diverse talent. We must continuously bring up the fact that scientific results show that more diverse teams are more creative and much more effective at research. As researchers, this research must appeal to us more than anyone else, right? Like, we know what makes research better, based on research.
A: Is there someone from an underserved/underrepresented group who really inspires you and/or has been instrumental in your success?
L: Am I allowed to name several? Several senior people like Kathyn McKinley, Sarita Adve, Margaret Martonosi inspire me. There are also many younger members of the community, like Michael Carbin, for example, whose work I really admire and respect tremendously.
K: José Moreira, was super supportive and really instrumental in shaping me as a researcher. I also got a lot of encouragement early on from Kathryn, Sarita, and Margaret. I am sure I am leaving many important people out— apologies in advance!
A: Wrapping up with my favorite question, what would you be doing for a living if you were not in computer science?
L: I’m interested in cooking, and food in general. I’m very market and season driven. For example, when it’s mushroom season, I cook a lot of mushrooms. We also spend a good chunk of our lives eating well. We like to explore different types of food and try to reproduce them or create new dishes by drawing inspiration from them. I also really enjoy cooking for other people.
K: I’m really into gardening, and this is another area where we’re complementary because I garden and he cooks. 🙂 I’d probably be a high-tech farmer of some sort who applies the latest technology to farming.
A: Thanks so much for your time. It was wonderful chatting with you!
L & K: Thank you for doing this!
Bio: Akshitha Sriraman is a Ph.D. candidate in Computer Science and Engineering at the University of Michigan. Her research bridges computer architecture and software systems, demonstrating the importance of that bridge in realizing efficient hyperscale web services via solutions that span the systems stack. Sriraman has been recognized with a Facebook Fellowship, a Rackham Merit Ph.D. Fellowship, and was selected for the Rising Stars in EECS Workshop. Her work has been recognized with an IEEE Micro Top Picks distinction.
She hopes to enter academia after her Ph.D. program, and is currently on the academic job market (for tenure-track faculty positions).
Disclaimer: This post was written by Akshitha Sriraman for the SIGOPS blog. Any views or opinions represented in this blog are personal, belong solely to the blog author and the person/people interviewed; they do not represent those of ACM SIGOPS, or their parent organization, ACM.