I am an Alvarez Fellow at Berkeley Lab and UC Berkeley. Previously, I completed my PhD at Stanford University, where I was supported by the Department of Energy Computational Science Graduate Fellowship. I also spent time at Los Alamos National Lab, Toyota Research Institute, and Google Research. Before this, I studied physics and chemistry in the College of Creative Studies at UC Santa Barbara.
My contact info is aditik1 at berkeley.edu.
I’m interested developing and utilizing methods in scientific machine learning, dynamical systems theory, numerical methods, and applied topology to study a range of science and engineering problems. Currently, I’m excited about exploring the coupling of domain-driven scientific mechanistic modeling with data-driven machine learning methodologies to accelerate and improve spatial and temporal modeling. Some examples of recent work include:
Characterizing the challenges associated with incorporating fundamental physical laws into the machine learning process (i.e., ‘‘physics-informed neural networks’’), and devising strategies to overcome their failure modes by changing the learning paradigm (Neural Information Processing Systems (NeurIPS), Arxiv version; 2021),
Learning interpretable and transferable scientific representations. For example, representing structure/symmetry in scientific systems by mapping data into topological descriptors (that are invariant to homeomorphic transformations of the domain). This includes learning and interpreting structure-property relationships for nanoporous materials such as zeolites (J. Phys. Chem. C.; 2020) and metal-organic frameworks (Sci. Reports; 2021), or coupling computational topology techniques with graph neural networks to learn protein/biomolecule structure-function relationships (NeurIPS Learning Meaningful Representations of Life Workshop; 2020),
Incorporating the “structure” of data into optimization algorithms (arXiv:2011.05290; 2020).
A full list of publications is available on Google scholar.