Year in Review: 2018
Summary: A copy of my end of year message to my lab, where I review the key events from the past year.
This past year the lab grew in a number of ways, both because existing projects that started at the end of 2017 have now hit full steam and because of new project funding that came in over the past year. Looking back with virtue of hindsight, I would characterize 2015 and 2016 as the lab’s “seedling period” where most of the groundwork for today’s projects was laid (e.g., Greg/Ryan/Jessica’s work on the MechProcessor, Wei’s work on Design Manifolds, Faez’s work on diverse ranking and matching) and 2017 was perhaps the “germination period” where we saw a number of projects take root and bring on new members. With that analogy, I consider 2018 the beginning of the “blooming period,” where we really got to see our hard work pay off with a series of exciting papers and new projects that have expanded the what our lab is capable of. Honestly, this has been the most intellectually fulfilling year of my career and a ton of fun! I have all of your hard work to thank for that. I am truly humbled to work with such a great group of researchers.
The best parts of my lab are the people, so I wanted to start my 2018 review by noting some of the important departures and arrivals this past year.
In January, Kailyn Cage (now Dr. Cage—the first member of her family to earn an PhD!) successfully defended her dissertation entitled “Optimizing Mass Customization through Consideration of Human Variability and Machine Specification Trade-Offs” and walked in the May, 2019 ceremony. Kailyn has since taken her immense talents to the SF Bay Area where she started in Amazon Lab 126 (Amazon’s hardware R&D lab located in Sunnyvale). We also bid farewell in May to two undergrad researchers (Josh Land, now a Software Eng. at Appian, and Ashwin Jeyaseelan) who help out in two ongoing DARPA projects. We were joined in the summer by high school interns Jaime Yen (Eleanor Roosevelt HS) and Isabel Beariault (Holton Arms HS) who were investigating materials for soft-robot design and locomotion. Jaime continues with us through May 2019 as part of ERHS’s research practicum (looking forward to the Science Fair next spring!). In the fall, we also welcomed Rachel Hess as a M.S. student working on our new NIH project and also Dr. David Anderson who joined us in November after finishing his Ph.D. under Kris Wood at SUTD. Lastly, in the next few weeks, Dr. Dan Elton, who has been with us since Summer 2017 will be taking up an exciting new position nearby as a Staff Scientist at the National Institutes of Health applying deep learning to medical imaging, primarily radiological images (CT Scans) and we will be joined by Dr. Jun Wang who just completed his Ph.D. under Rahul Rai at UBuffalo.
I am continually impressed by the range of backgrounds all of you bring to the table—from mechanical engineering to physics to statistics/applied math to computer science. Many of the ideas we have been exploring together involve the unique intersection of your many fields; for example, connections between differential geometry and generative models, or between type theory and conceptual design, among many others. Such bridges would simply not be practical without you. You really do make the whole greater than the sum of their parts! This expansion outgrew our old lab in the Engineering Lab building, and led to a new lab space in Martin Hall that we are still in the process of renovating. I’m looking forward to building out that space in 2019!
On a related note personal to me, 2018 saw my undergrad Gemstone team graduate from UMD (Team PRINT). While they didn’t spend much time seated in the lab with you all, some of them interacted with you at lab social events and they have been a big part of my life at UMD for the past four years. (Gemstone is a four-year intensive research program where I advised them during their time at UMD.) This culminated with them successfully defending their undergrad thesis and also getting a paper accepted in and presented at IDETC 2018, which I consider impressive considering they were all undergrads not working with a grad student or post-doc. Many of them are off to high-profile jobs or graduate schools, including Microsoft, UC Berkeley, NAVAIR, Textron, UCLA, among others.
In broad terms, your scientific efforts have led to our most productive year to-date, by almost every measure I can think of. Depending on how exactly you count it, this year we published around 15 papers, more than double that of last year. This includes multiple papers in the flagship journals of my field (including JMD and SAMO), largely thanks to the excellent work of Faez and Wei. They have consistently put forward brilliant work that I think really drives at some mathematical boundaries of my “home” field in unique ways.
Perhaps just as important, however, we also expanded the reach of our work into broader readership journals like Scientific Reports (part of the Nature publishing group) and Machine Learning venues (like ICML and NeurIPS). Here thanks is due to Dan and Zois for their ambition and drive to get our work out there into related fields that lie beyond my traditional publishing community; I have learned a great deal from both of you. We’ve also begun making inroads at bridging design and robotics, particularly via Charlie’s collaborations with Sarah Bergbrieter on soft robotic actuators that he presented at the IROS soft robotics workshop in the fall. Likewise, a collaboration between Wei and Nicholas has resulted in an invited AIAA SciTech submission; I am looking forward to seeing how our work might integrate with the efforts of the aerospace community. I’m enjoying watching all the ways that you are all making design’s contributions more relevant to other fields—it is only through such efforts that we’ll ultimately expand the impact of design to many more areas of society and get out of our academic bubble.
While papers are easy to measure, your efforts have also led to scientific advances in other ways that often more difficult to keep track of. For example, earlier this year, Charlie, Nicholas, and Josh worked immensely hard at the tail end of Phase I of the DARPA Fundamental Design program. Their efforts were rewarded since we were awarded the Phase II funding for the program over some other very deserving and top-notch institutions and companies. Because of their hard work, I have had the pleasure of being invited to give talks at two NASA laboratories (Langley and Goddard), ARPA-E, and United Technologies all of which were interested in building upon our efforts. Charlie, Nicholas, and now David are exploring some honestly foundational work at the core of ML + Design that we are in the process of writing up for 2019. Even if they are only partially successful, they could fundamentally re-envision the basis for computational design in ways that really build on our strengths. If you haven’t had the chance to chat with them in lab about what they are working on, I encourage you to do so.
In terms of new projects starting up, Faez’s work on diversity algorithms last year laid the foundation for a new NSF grant we just received that will explore how idea diversity impacts final solution quality. Wei’s airfoil manifold papers led to the preliminary results that we used to support a new NIH grant on 3D printable grafts for children with congenital heart disease (this is the project Rachel is now on). Charlie and Nicholas’ work via our FUN DESIGN project provided background material that led us to win a Generate Design project with Lockheed Martin over the next year. While I don’t bother many of you with these kinds of details on a day-to-day basis, I think it is important that you see how the hard work and collaborations you set up really do impact the lab’s activities in the long term.
While there has generally been renewed interest in the combination of ML and Mechanical Design for the past decade or so, it does seem to me like 2018 saw a marked increase the pace and scale of this interest. For example, in the later part of this year, the Journal of Mechanical Design launched a special issue on the topic of Machine Learning and Engineering Design, and then Design Science followed closely thereafter with a special issue on Deep Learning. At ASME IDETC, machine learning tended to be a specialized topic and was even considered an “emerging topic” in DAC as of only 3-4 years ago. Now, by contrast, sessions with significant coverage of Machine Learning occur in all major design conferences, both as full sessions as well as workshops (e.g., I hosted a Deep Learning workshop at Design Computing and Cognition this past summer). Even in venues not specifically targeting design’s intersection with machine learning, we’ve been seeing pickup; for example, in the Special Issue on Design Creativity that Kate Fu, Dave Brown, and I guest edited for Artificial Intelligence for Engineering Design Analysis and Manufacturing (AIEDAM), ML approaches played a part.
This has also been mirrored in a growing willingness by federal and industry funders to explore ML + Design approaches in a variety of domains. As one example, if you take a look at DARPA’s AI Next Campaign (the agency’s current AI portfolio), of the roughly twenty programs listed there two of them are design programs (that’s around 10% of the AI programs). If you had told me five years ago that that would be the case today, I would not have believed you. Interests do tend to go in cycles and I’m not sure how long such trends will last, but for the time being it is certainly refreshing to see academic, industry, and government sectors all giving the intersection of ML and Design a chance. It is now up to us to make sure we deliver the kind of results that demonstrate design’s value, both within and beyond our field!
Occasionally, I’ll get some updates from past lab alumni. Over this past year, we had a visit from Greg Carmean, who is still nearby at NSWC Carderock. We had fun looking at how the campus has changed since he was here and discussed how he is helping modernize how his department at NSWC-CD handles things like code review and version control (yay Git!).
I also talked with Ceena Modarres who is still enjoying being up in NYC as a Data Scientist at Capital One after completing the Data Incubator program here in DC. His group there has been doing some cool collaborations with John Paisley’s group at Columbia, and he was actually at NeurIPS a few weeks ago presenting at the AI in Financial Services workshop on some work in explainable deep learning for credit lending.
Both Greg and Ceena said to say hi to everyone!
Thanks everyone again for a great 2018, and I’m looking forward to what we can accomplish together in 2019!