My passion lies in developing methods and codes for modeling electronic structure in strongly correlated and complex materials for phenomena and processes such as light-matter interactions, adsorbate-substrate interactions, defects, and thin-film deposition. During my Ph.D., I focused on improving and accelerating quantum embedding techniques for molecular and periodic systems. I implemented DFT-based embedding methods and coupled them with RT-TDDFT, many-body methods (GW/BSE), and correlated wavefunction methods (MP2 and CCSD). This allowed obtaining highly accurate ground-state and excited-state properties of complex materials efficiently at reduced computational cost. Notably, my work also allows simulating nonlinear optical phenomena like high harmonic generation for large systems. Currently, I am developing methods that allow simulating materials efficiently using quantum computing algorithms such as the variational quantum eigensolver (VQE).
In my postdoctoral research, I have diversified into the development of machine-learned interatomic potentials (MLIPs) and MLIP-accelerated workflows to study chemical vapor deposition (CVD) and physical vapor deposition (PVD) of thin films on substrates. To explore deposition processes at the atomic scale, I use DFT calculations combined with nudged elastic band (NEB) methods to study reaction pathways and energy barriers, as well as ab initio molecular dynamics (AIMD) to study dynamic evolution. Using the high-fidelity data from the DFT simulations, I develop efficient MLIPs and fine-tune foundation models to extend their applicability.
The methods developed by me are available in widely used electronic-structure and materials-simulation software. I am one of the lead developers of TURBOMOLE, an efficient electronic-structure program cited by thousands worldwide, and can perform hybrid-functional periodic DFT calculations for systems with approximately 15,000 electrons (about 2,700 atoms). I am also the founding developer of PyFock, the first Indian Gaussian-basis DFT code with GPU acceleration for large molecular systems, CrysX, and MLIP Studio, a platform for benchmarking and applying machine-learned interatomic potentials to materials problems. Together, these efforts reflect my broader goal of making advanced electronic-structure and atomistic-simulation methods more accurate, efficient, and accessible to the scientific community.
Outside research, I enjoy creating YouTube tutorials, web and Android apps, and computer software and libraries for researchers and students.