Profile picture
Administrative data

Rohit Goswami

rohit.goswami@epfl.ch +41 21 693 77 17 https://rgoswami.me

Expertise

Physical Chemistry & Statistical Mechanics: Specialized in molecular dynamics (MD), transition path theory, and rare event sampling.

Computational Materials Science: Expert in density functional theory (DFT) and itinerant electron magnetic interactions.

Scientific Software Engineering: Specialist in high-performance computing (HPC), Fortran-Python interoperability, and the Nix ecosystem.

Machine Learning for Chemistry: Developer of Bayesian analysis tools and Gaussian Process Regression (GPR) models for potential energy surface (PES) saddle searches.

Mission

I strive to advance the reproducibility of computational research through robust software ecosystems. My work bridges the gap between theoretical physical chemistry and scalable high-performance computing. By developing standardized APIs and maintainable open-source tools, I enable researchers to simulate complex atomic systems with greater accuracy and efficiency. In particular, I have a focus on long timescale dynamics of systems using adaptive kinetic monte carlo.

Current Work

At EPFL, Rohit focuses on the metatensor ecosystem within the COSMO lab under Prof. Michele Ceriotti. This work involves creating modular, high-performance components for atomistic machine learning. Additionally, he scales off-lattice object kinetic Monte Carlo (KMC) simulations to millions of atoms, specifically investigating heating effects in large copper systems. He continues to maintain f2py for NumPy, ensuring the longevity of Fortran-based scientific code in modern Python environments.

Rohit Goswami currently serves as a Systems Specialist at the Laboratory of Computational Science and Modelling (COSMO) at EPFL. His academic journey began with a B.Tech. in Chemical Engineering from Harcourt Butler Technical University, followed by doctoral research at the University of Iceland. His doctoral project, funded by the Icelandic Research Fund, focused on modeling magnetic interactions of itinerant electrons which led to his dissertation on efficient representations for chemical kinetics using Bayesian machine learning.

Throughout his career, Rohit has maintained a deep commitment to the scientific software community. He acts as an Editor for the Journal of Open Source Software (JOSS) and contributes significantly to foundational libraries like NumPy. His diverse experience includes visiting research positions at Queen's University and industrial software engineering roles at Quansight. He has been elected to membership of the Institute of Physics (IOP), Royal Society for Chemistry (RSC), and the Chartered Institute for IT (BCS).