Anjali Kumari Participates in Joint ICTP–MARVEL College on Materials Simulations in the Age of AI

The two-week intensive college brought together early-career researchers to explore modern approaches to electronic-structure and atomistic simulations, combining foundational methods such as first-principles calculations and molecular dynamics with emerging machine-learning techniques. The program included lectures as well as hands-on computational sessions.

Anjali received a scholarship from the organizers to attend the conference. During the event, she presented her poster, “Accelerating Optoelectronic Device Design through Machine Learning–Driven Multi-Output Regression and Ab-Initio Model in Double Perovskites,” co-authored with Anup Shrivastava and Jost Adam.

Her work presents an integrated machine-learning and density-functional-theory (DFT) framework for accelerating the exploration of double perovskites for optoelectronic applications. In particular, the study employs multi-output regression to simultaneously predict bandgap and dielectric constant across a large compositional space. Among the five machine-learning approaches investigated, XGBoost achieved the best overall performance, and the predictions were subsequently validated against DFT calculations. The results show the potential of combining machine learning with first-principles methods to enable scalable, physically reliable materials screening.

Congratulations to Anjali on receiving the scholarship and on successfully representing the CMP group at this international training program!