Atharva Kelkar

Position title: Graduate Student


Phone: he/him/his

Room 2029, 1415 Engineering Drive, Madison, WI 53706



Atharva was born and raised in Thane, India. He earned a dual Bachelor’s and Master’s degree in Chemical Engineering from the Indian Institute of Techonology Bombay in 2016, post which he worked for 2 years with A.T.Kearney in their Mumbai office. Atharva joined the Van Lehn group in the Fall of 2018.


The advent of machine learning techniques has opened up exciting opportunities to analyze molecular dynamics data, which can provide insights based on far fewer computations than are currently required. Neural networks, with their ability to identify spatial correlations over multiple length scales, provide a great way to analyze physical data. My research is focused on training neural networks to learn the characteristics of hydrophobic surfaces, specifically in self-assembled monolayer protected surfaces. These trained networks can then be used to (1) rapidly screen surfaces for hydrophobicity and guide experiments (2) to learn the manifestation of the underlying physics of the system from these networks based on their maximally activating features.


A. S. Kelkar, B. C. Dallin, and R. C. Van Lehn. “Predicting Interfacial Hydrophobicity by Learning Spatiotemporal Features of Interfacial Water Structure: Combining Molecular Dynamics Simulations with Convolutional Neural Networks.” Journal of Physical Chemistry B2020, 124 (41), 9103-9114. [Link]