Spandan Das: Self-Taught Machine Learning Intern, NCCS User, College Student, and Published Author

Hometown: I used to be born in Milwaukee, Wisconsin, however I’ve lived most of my life in Fairfax, Virginia.

What was your profession path to NASA? I had studied pc science in highschool on the Thomas Jefferson Excessive Faculty for Science and Know-how (TJHSST) in Alexandria, Virginia, together with a course in machine studying. I additionally led the highschool pc group and studied pc science by myself by way of on-line programs and in programming contests. A couple of months previous to the beginning of my summer season 2020 internship, throughout COVID-19 quarantine restrictions, I made a decision to review machine studying by myself, primarily by way of self-study utilizing books and programs supplied on-line. That preparation and data of pc science served me nicely for my utility for a NASA internship. I ultimately accomplished two priceless internships in summer season 2020 and 2021 on the NASA Goddard House Flight Heart.

An architectural rendering of the primary entrance of the Thomas Jefferson Excessive Faculty for Science and Know-how by the Architect of Report for the campus-wide renovation, Ballou Justice Upton Architects. Rendering courtesy of BJU Architects.

Throughout my first internship in summer season 2020, I used to be a rising highschool senior. I used to be matched with my internship mentor, Dr. Jie Gong, a analysis atmospheric scientist within the Climate and Radiation Lab at NASA Goddard, primarily based on my background in pc science and machine studying. Initially, Dr. Gong and I met remotely and mentioned what I might be engaged on that first summer season. An necessary requirement for one in every of her personal main tasks was the power to foretell the kind of classification of an atmospheric phenomena (precipitation, on this case), primarily based on metadata supplied by NASA’s International Precipitation Measurement mission (GPM) satellites. I began engaged on that undertaking in summer season 2020 and made good progress on utilizing machine studying and testing precipitation fashions. I in contrast outcomes utilizing knowledge from two devices, one with an energetic sensor, which is dearer to construct and preserve, and one with a passive sensor, which is inexpensive. I offered my outcomes to Dr. Gong’s group and to the Local weather and Radiation Lab on the finish of my first summer season internship — remotely after all, due to COVID lockdown. My mentor appreciated my work, and I appreciated the undertaking.

After graduating highschool, I had a second summer season internship with Dr. Gong at NASA Goddard in 2021, additionally working remotely. I continued engaged on the outcomes from the analysis begun the earlier summer season, testing extra complicated machine studying fashions and increasing on and decoding these earlier findings. The mixed outcomes of these two summers of labor culminated in a coauthored paper printed in July of 2022 within the skilled journal, Distant Sensing.

How did having to work remotely as an intern and the expertise of working with Dr. Gong influence you, when it comes to your future analysis pursuits? The analysis was, after all extraordinarily rewarding in itself. Earlier than my internships, I had by no means labored with or attended conferences with scientists. Throughout my internship, I met with my mentor every day. I used to be additionally in a position to attend common conferences with a bunch of 6-7 researchers. At first, I used to be very intimidated, however I used to be in a position to finally be snug with asking questions. Having the chance to work together with these NASA scientists frequently and ask questions was one of many coolest components of my internship expertise. I used to be initially deeply centered on tackling this undertaking from a pc science and machine studying perspective. However at these conferences with NASA specialists of their fields, I used to be in a position to study extra about what the information and outcomes meant, and so they had been in a position to level me in the precise route. They prompt different options so as to add which may assist the mannequin and shared their concepts on which fashions may work higher as a result of sure knowledge is distributed a sure method, which was eye-opening. I received suggestions from the top of the lab after my department-wide presentation was additionally useful. That each one made for an important expertise.

How will this analysis assist enhance or influence future fashions? The influence of this undertaking is that we’re ready to make use of knowledge from a passive sensor, plug that knowledge right into a machine studying mannequin, and get precipitation outcomes comparable knowledge to that from an energetic sensor. Utilizing knowledge from the inexpensive passive sensor with machine studying means fewer sensors, finally saving NASA prices for the following era of satellites and distant sensing devices used for precipitation observations.

One main problem for the International Local weather Mannequin (GCM),” defined Dr. Gong, “is to accurately partition convective and stratiform precipitation processes, which immediately impacts the mannequin’s illustration of the power stability and hydrological cycle. With the assistance of machine studying fashions and the computational energy supplied by NCCS, we will use a number of spaceborne passive sensors (e.g., GPM-GMI) skilled by one spaceborne energetic sensor (GPM-DPR on this case) to trace the spatial and temporal evolutions of precipitation programs and their buildings. This can’t solely additional our data of understanding how precipitation types in numerous stage and climate system regimes, however also can assist consider and enhance GCMs.

How did the NASA Heart for Local weather Simulation (NCCS) help this analysis? NCCS supplied me with ample compute energy to construct, prepare, and check varied studying fashions which had been essential to my undertaking as an intern in addition to the analysis that ensued. I merely couldn’t have run and check any of those simulations on desktops or laptops with out the GPUs and supercomputing assets of the NCCS. “The NCCS help group is excellent,” emphasised Dr. Gong. “They’re extraordinarily supportive {and professional}.”

What did you do after your summer season internships ended? After graduating highschool in spring 2021 and finishing my second NASA summer season internship, I moved to Pittsburgh, Pennsylvania in fall 2021 and began engaged on an undergraduate diploma at Carnegie Mellon College (CMU). I’m now in my second 12 months at CMU, learning pc science with a focus in machine studying. My main analysis pursuits embody exploring the functions of machine studying to robotics, imaginative and prescient, and finance.

What or who particularly evokes you? My lecturers and mentors all the time encourage me and push me to do extra. Two specifically have been extraordinarily influential in creating my curiosity in utilizing expertise to resolve difficult and significant issues.

In highschool, Mr. Malcolm Eckel used the Socratic technique to show members of his synthetic intelligence course how one can arrange an issue utilizing the instruments that we had. He helped us perceive the relative utility of present algorithms and how one can develop new ones. This all helped us totally perceive the ground-up method to issues and perceive and give attention to why we had been doing what we had been doing.

Dr. Jie Gong, my NASA internship mentor for 2 summers, impressed and inspired me. I had by no means labored one-to-one with a scientist like that, and he or she spent loads of time with me. I used to be lucky in some ways to have Dr. Gong’s mentoring. She met with me every day, regardless of the pandemic and the calls for of conducting her personal analysis, collaborating nearly. She was centered on guaranteeing that I discovered new issues over the summer season and inspired me to publish my outcomes in order that I might study in regards to the scientific analysis course of from begin to end. Dr. Gong additionally made positive that I had loads of networking alternatives, regardless of the quarantine limitations. She invited me to nearly attend weekly Local weather and Radiation division conferences, the place I used to be in a position to get priceless, first-hand suggestions on my undertaking from numerous NASA researchers. As well as, she gave me the chance to go to NASA Goddard for an in-person networking alternative, the place I met different interns and NASA Administrator Invoice Nelson.

Spandan Das, at left, with NASA Administrator Invoice Nelson at middle, different interns and NASA employees at a networking assembly for interns in summer season 2021 at NASA Goddard.

Are there any folks in your subject who’ve influenced you? One one who all the time evokes me is my father, who labored for a few years throughout my childhood to finish his Ph.D. in pc science on the College of Wisconsin Milwaukee whereas additionally elevating our household. Seeing my father research complicated matters and develop new algorithms, even after lengthy hours at work, has impressed me to push by way of tough occasions, each inside and out of doors of analysis.

What challenges have you ever needed to overcome? One latest problem that I’ve needed to overcome was balancing my faculty’s course load with analysis. Whereas this mix has led to lengthy hours, it has been extraordinarily rewarding in my growth as a scholar and analysis scientist.

When it comes to scientific analysis and your schooling, the place are you heading subsequent? I’m nonetheless figuring that out, however I feel {that a} graduate diploma is probably going.

Associated Hyperlink

• Das, S., Y. Wang, J. Gong, L. Ding, S.J. Munchak, C. Wang, D.L. Wu, L. Liao, W.S. Olson, and D.O. Barahona, 2022: A Comprehensive Machine Learning Study to Classify Precipitation Type over Land from Global Precipitation Measurement Microwave Imager (GPM-GMI) Measurements. Distant Sensing, 14, 3631, doi:10.3390/rs14153631.

Sean Keefe, NASA Goddard House Flight Heart
November 30, 2022