Shreyas Gaikwad

Graduate Research Assistant

I am a Ph.D. candidate in the Computational Science, Mathematics, and Engineering (CSEM) program at The Oden Institute for Computational Engineering and Sciences at The University of Texas (UT) at Austin. I am advised by Dr. Patrick Heimbach as part of the research group CRIOS. I am broadly interested in data assimilation under uncertainty, machine learning, optimization, and mathematical modeling of systems.

My research primarily involves PDE-constrained inverse problems in the context of paleoclimate reconstruction over ice sheets. We are trying to leverage the recently collected radiostratigraphy-based age layer data to constrain the paleoclimate of the ice sheet during the last 11,000 years. We hypothesize that a better-constrained ice sheet model can better forecast sea level rise contributions in the next century.

I have also been exploring the utility of Explainable AI (XAI) methods such as Layerwise Relevance Propagation (LRP) and how they compare to adjoint gradients in an oceanic and glaciological context.

I am a reviewer at the Journal of Open Source Software (JOSS) and the Journal of Mountain Science (JMS). I am passionate about Open Source Software Development and have contributed substantially to efforts to make open-source adjoint models available for data assimilation efforts of the glaciology and oceanography communities by leveraging open-source Automatic Differentiation (AD) tools. I am the lead developer and maintainer of the ice-sheet model adjoint SICOPOLIS-AD v2. We have developed an open-source adjoint for the ocean general circulation model, MITgcm-AD, which offers an alternative to a proprietary version that costs ~$14,000 per year per individual license. I have also developed Julia codes which interface Enzyme and MPI for glacier models and reported fundamental bugs within Enzyme.

Over the course of my work, I have collaborated with researchers from Indian Institute of Technology (IIT) Bombay, UT Austin, Argonne National Laboratory, Scripps Institution of Oceanography, Hokkaido University, IIT Kanpur, MIT, and Institut National de Recherche en Informatique et Automatique(INRIA), France.

Previously, I completed a B.Tech in Mechanical Engineering (with Honors) and a Minor in Computer Science from the IIT Bombay, Mumbai, India in May 2019. I received my Master’s in Computational Science, Engineering, and Mathematics from UT-Austin in May 2021.

In my free time, I like hiking, playing chess, cricket, and reading (mostly history and fiction). I have also recently taken up cubing, and can solve up to a 5x5. All of these activities are a lot more fun when Baloo is also involved.

News


UT STEM Girl Day, 2024 : Helping organize and demonstrate to elementary and middle school kids how to conduct simple yet powerful experiments to understand the role of sea ice and glacial melts in the changing Arctic climate and explore observational and modeling tools fundamental to many CRIOS research projects.


Our work on MITgcm-AD v2 is submitted to Joint Laboratory on Extreme Scale Computing Future Generation Computer Systems (JLESC-FGCS), 2024. Check out the preprint on arXiv!


I will be giving a talk on “Computational Science To Enable Digital Twins Of The Ocean” at minisymposium “Mathematical and computational foundations of predictive digital twins” at the 6th Annual Meeting of the SIAM Texas-Louisiana Section (TXLA23) to be held at the University of Louisiana at Lafayette, Lafayette, Louisiana.


I am going to Qeqertarsuaq, Greenland in August-Septemeber 2023 as part of the prestigious ACDC/GRISO summer school 2023! We will be based at the Arctic Station there.