Hi, I’m Dakshesh! I’m a PhD researcher at the UCL Data Intensive Science CDT using graph-based machine learning methods to study how galaxy properties are influenced by their large-scale structure environments. My research combined deep learning, statistics, and astrophysics for a comprehensive understanding of galaxy evolution.

Research Interests

• Graph ML and geometric learning for cosmology: Building graph-based representations of the galaxy distribution that preserve local geometry and physically meaningful structure.
• Dark matter environment inference: Classifying galaxies into cosmic web environments (voids, walls, filaments, clusters) using simulation-grounded machine learning.
• Galaxy evolution in context: Quantifying how galaxy properties depend on large-scale structure and environmental history.
• Simulation-to-observation transfer: Adapting methods trained on simulations toward survey applications, including DESI-era data analysis.
• Interpretable scientific machine learning: Designing models whose behavior can be connected back to astrophysical mechanisms, not only predictive performance.

Latest Paper

Learning the Cosmic Web: Graph-based Classification of Simulated Galaxies by their Dark Matter Environments
Dakshesh Kololgi, Krishna Naidoo, Amelie Saintonge, Ofer Lahav

This work presents a three-step approach linking simulated galaxies to the dark matter cosmic web: T-Web labels of the matter field, Delaunay graph metrics for each galaxy, and graph-attention classification. For stellar mass-selected galaxies ((M_* > 10^9 M_\odot)), the GAT+ model reaches 85% accuracy and outperforms graph-agnostic and graph-convolution baselines.

Read here: arXiv:2512.05909

CV

Here’s a preview of my latest CV.