June 28, 2022 — Mansi Sakarvadia, a University of North Carolina at Chapel Hill student who works as a research assistant at the Argonne Leadership Computing Facility (ALCF), was recently named one of the US Department of Energy’s (DOE) 33 recipients ) prestigious Computational Science Graduate Fellowship (CSGF). The ALCF is a user facility of the DOE Office of Science at Argonne National Laboratory.
As a DOE CSGF fellow, Sakarvadia can continue her work in applying high-performance computing (HPC) to computer science research and connect with fellow researchers to help realize her thesis.
Last summer, Sakarvadia took part in the ALCF’s Summer Internship Program for Students, working on a project optimizing deep learning applications for supercomputers. This year she returned to the ALCF to work as a research assistant in the data science group.
In this Q&A, Sakarvadia discusses what fueled her interest in scientific informatics, her current research projects and her future ambitions.
Can you start by telling us what inspired you to pursue a career in computational science?
I had a most enjoyable and formative experience completing a Summer Undergraduate Learning Internship (SULI) in 2021 at the Argonne Leadership Computing Facility (ALCF). The project I was working on that summer (FLOPs Aware Deep Learning) captivated me and helped me realize that I wanted to continue working on similar issues in the future. The project experience, as well as the mentorship I received from my supervisors, Kyle Felker and Taylor Childers (and the entire ALCF data science team), helped me realize how many exciting research questions exist within the broad field of computational science. My positive experiences at Argonne, as well as several computational science research projects that I had been involved in at my undergraduate institution, eventually sealed the deal and inspired me to pursue a computational science-focused research career.
What kind of work have you done at the ALCF?
Before my SULI appointment in 2021, my “FLOPs Aware Deep Learning” project focused on exploring methods to design computationally cost-conscious neural networks that automatically maximize the throughput of the target hardware, given a specific set of deep learning frameworks and libraries . After the position ended, I spoke with my summer mentors and decided to return to the ALCF as a research assistant to work on additional projects broadly related to the benchmarking and integration of Vision Transformers (a type of neural network architecture) into scientific workflows. . I was able to gain a lot of experience with the ALCF’s ThetaGPU machine, exploring the lowest architectural details of the A100 GPUs and training neural networks across dozens of devices at once.
What attracted you to apply for the DOE Computational Science Graduate Fellowship?
I initially heard about the DOE Computational Science Graduate Fellowship from my ALCF mentor, Kyle Felker, who is an alumnus of the fellowship. After doing some additional research on my own, I decided that the goals of the fellowship matched mine. I was especially excited to apply for the fellowship because recipients at some point during their Ph.D. have to conduct a research practical in a DOE laboratory. studies on a topic unrelated to their dissertation. Because of my positive experience at Argonne, I felt that this opportunity would be a unique way to explore additional interdisciplinary research areas that expand my field of knowledge.
How will your experience at the ALCF affect your work with the CSGF program?
My time at ALCF initially exposed me to high-performance computing and challenged me to think about the types of problems associated with computing on large systems. Being a key focus of the CSGF program, I feel that my ALCF experience has equipped me with fundamental tools to conduct research in this area.
Can you tell us which areas of research you would like to focus on with your fellowship?
In general, I’m interested in machine learning and high-performance computing, and would love to work on some problems that are at the intersection of the two fields. Currently, I hope to apply research methods to build scalable, privacy-protecting machine learning models for high-performance scientific workflows. I am also interested in environmental science and would like to apply my computer science research to environmental science applications.
What aspect(s) of the fellowship are you most looking forward to and what do you hope to get out of the program?
I am most excited about participating in the practicum – I hope the experience will provide skills that complement my thesis work, but also broaden my research focus. I also look forward to building a network with DOE researchers, current fellows and CSGF alumni. It is my hope that this experience will allow me to remain involved in DOE research throughout my career.
The Argonne Leadership Computing Facility provides supercomputing capabilities to the scientific and engineering community to advance fundamental discovery and understanding across a wide range of disciplines. The ALCF is supported by the Office of Science, Advanced Scientific Computing Research (ASCR) program of the U.S. Department of Energy (DOE) and is one of two DOE Leadership Computing Facilities in the country dedicated to open science.
Argonne National Laboratory seeks solutions to pressing national problems in science and technology. Argonne, the country’s first national laboratory, conducts industry-leading basic and applied scientific research in virtually every scientific discipline. Argonne’s researchers work closely with researchers from hundreds of companies, universities, and federal, state and municipal agencies to help them solve their specific problems, advance America’s scientific leadership, and prepare the nation for a better future. future. With employees from more than 60 countries, Argonne is led by: UChicago Argonne, LLC for the Office of Science of the United States Department of Energy†
The Office of Science of the United States Department of Energy is the leading proponent of basic science research in the United States, working to address some of the most pressing challenges of our time. For more information visit https://energy.gov/science†
Source: Logan Ludwig, ALCF