Spandan Das: Self-taught machine learning intern, NCCS user, student, and published author

Spandan Das: Self-taught machine learning intern, NCCS user, student, and published author

Spandan Das: self-taught machine learning intern,
NCCS user, student and published author

Hometown: I was born in Milwaukee, Wisconsin, but have lived most of my life in Fairfax, Virginia.

What was your professional background at NASA? I had studied computer science in high school at Thomas Jefferson High School for Science and Technology (TJHSST) in Alexandria, Virginia, including a course in machine learning. I also led the high school IT team and studied IT on my own through online courses and programming competitions. A few months before the start of my 2020 summer internship, during quarantine restrictions related to COVID-19, I decided to study machine learning on my own, mainly self-taught using books and courses offered online. This preparation and knowledge of computer science served me well for my application for an internship at NASA. I finally completed two valuable internships in the summer of 2020 and 2021 at NASA Goddard Space Flight Center.

An architectural rendering of the main entrance to Thomas Jefferson High School for Science and Technology by the architect responsible for the campus-wide renovation, Ballou Justice Upton Architects. Rendered courtesy of BJU Architects.

During my first internship in the summer of 2020, I was a rising high school student. I was matched with my internship mentor, Dr. Jie Gong, an atmospheric science researcher at NASA Goddard’s Climate and Radiation Lab, based on my background in computer science and machine learning. Initially, Dr. Gong and I met remotely and discussed what I would be working on this first summer. An important requirement for one of his own major projects was the ability to predict the classification type of an atmospheric phenomenon (precipitation, in this case), based on metadata provided by the satellites of the Global Precipitation Measurement mission. (GPM) from NASA. I started working on this project in the summer of 2020 and have made good progress in using machine learning and testing precipitation models. I compared the results using data from two instruments, one with an active sensor, which is more expensive to build and maintain, and the other with a passive sensor, which is less expensive. I presented my results to Dr. Gong’s team and the Climate and Radiation Lab at the end of my first summer internship – remotely of course, due to the COVID lockdown. My mentor liked my work and I liked the project.

After graduating from high school, I completed a second summer internship with Dr. Gong at NASA Goddard in 2021, also working remotely. I continued to work on the results of the research started the previous summer, testing more complex machine learning models and expanding and interpreting these earlier findings. The combined results of these two summers of work resulted in a co-authored article published in July 2022 in the professional journal, remote sensing.

Dr. Jie Gong, atmospheric scientist at NASA Goddard’s Climate and Radiation Branch.

How has working remotely as an intern and working experience with Dr. Gong impacted you, in terms of your future research interests? The research was, of course, extremely rewarding in itself. Before my internships, I had never worked or participated in meetings with scientists. During my internship, I met my mentor daily. I was also able to attend regular meetings with a group of 6-7 researchers. At first I was very intimidated, but I ended up being comfortable asking questions. Having the opportunity to regularly interact with these NASA scientists and ask questions was one of the most interesting parts of my internship experience. Initially, I was deeply focused on how to approach this project from a computational and machine learning perspective. But in these meetings with NASA experts in their fields, I was able to learn more about the meaning of the data and the results, and they were able to point me in the right direction. They suggested other features to add that might help the model, and shared their thoughts on which models might work better because some data is distributed in a certain way, which was an eye opener. I received feedback from the head of the lab after my department-wide presentation was also helpful. It all made for a great experience.

How will this research help improve or influence future models? The impact of this project is that we are able to take data from a passive sensor, plug that data into a machine learning model, and get rainfall results similar to an active sensor. . Using data from the less expensive passive sensor with machine learning means fewer sensors, ultimately saving NASA costs for the next generation of satellites and remote sensing instruments used for precipitation observations.

A major challenge for the Global Climate Model (GCM),” Dr. Gong explained, “is to properly partition convective and stratiform precipitation processes, which has a direct impact on the representation of the energy balance and the hydrological cycle by the model. With the help of machine learning models and the computing power provided by the NCCS, we can use multiple spatial passive sensors (e.g. GPM-GMI) trained by a spatial active sensor (GPM-DPR in this case) to follow the spatial and temporal evolutions. precipitation systems and their structures. This can not only deepen our knowledge of how precipitation forms in different stage and weather system regimes, but can also help in evaluating and improving GCMs.

How did the NASA Climate Simulation Center (NCCS) support this research? NCCS provided me with enough computing power to build, train and test various learning models that were crucial for my project as an intern as well as for the research that followed. I simply could not have run and tested any of these simulations on desktop or laptop computers without the GPUs and supercomputing resources of NCCS. “The NCCS support team is superb,” ​​Dr. Gong pointed out. “They are extremely supportive and professional.”

What did you do after your summer internships ended? After graduating from high school in the spring of 2021 and completing my second summer internship at NASA, I moved to Pittsburgh, PA in the fall of 2021 and began working on a graduate degree. undergraduate at Carnegie Mellon University (CMU). I am now in my second year at CMU, studying computer science with a concentration in machine learning. My main research interests include exploring the applications of machine learning to robotics, vision, and finance.

What or who particularly inspires you? My teachers and mentors always inspire me and push me to do more. Two in particular were hugely influential in developing my interest in using technology to solve difficult and meaningful problems.

In high school, Mr. Malcolm Eckel used the Socratic Method to teach members of his Artificial Intelligence class how to set up a problem using the tools we had. It helped us to understand the relative usefulness of existing algorithms and to develop new ones. All of this helped us get a good grasp of the basic approach to the issues and to understand and focus on why we were doing what we were doing.

Malcolm Eckel, math teacher at Thomas Jefferson High School for Science and Technology.

Dr. Jie Gong, my internship mentor at NASA for two summers, inspired and encouraged me. I had never worked one-on-one with a scientist like that, and she spent a lot of time with me. I was fortunate in many ways to have Dr. Gong’s mentorship. She met with me daily, despite the pandemic and the demands of conducting her own research, collaborating virtually. She made an effort to ensure that I learned new things over the summer and encouraged me to publish my results so that I could learn about the scientific research process from start to finish. Dr. Gong also made sure I had plenty of networking opportunities, despite quarantine restrictions. She invited me to attend the weekly Climate and Radiation Department meetings virtually, where I was able to get valuable first-hand feedback on my project from a number of NASA researchers. Additionally, she gave me the opportunity to go to NASA Goddard for an in-person networking opportunity, where I met fellow interns and NASA Administrator Bill Nelson.

Spandan Das, left, with NASA Administrator Bill Nelson center, other interns and NASA staff during a summer 2021 intern networking meeting at NASA Goddard.

Are there people in your field who have influenced you? A person who always inspires me is my father, who worked for many years during my childhood to complete his doctorate. in computer science at the University of Wisconsin in Milwaukee while raising our family. Watching my father study complex topics and develop new algorithms, even after long hours of work, inspired me to get through tough times, both inside and outside of research.

What challenges did you have to overcome? A recent challenge I had to overcome was balancing my school’s course load with research. Although this combination resulted in long hours, it was extremely rewarding in my development as a student and researcher.

In terms of scientific research and training, where are you going next? I’m still figuring this out, but I think a graduate degree is likely.

Related link

  • Das, S., Y. Wang, J. Gong, L. Ding, SJ Munchak, C. Wang, DL Wu, L. Liao, WS Olson, and DO Barahona, 2022: A Comprehensive Machine Learning Study to Classify Types of precipitation on Earth from measurements of the Global Precipitation Measurement Microwave Imager (GPM-GMI). remote sensing, 143631, doi: 10.3390/rs14153631.

Sean Keefe, NASA Goddard Space Flight Center
November 30, 2022

#Spandan #Das #Selftaught #machine #learning #intern #NCCS #user #student #published #author

Leave a Comment

Your email address will not be published. Required fields are marked *