Research
My research is focused on radio observations of astronomical transients, mainly Tidal Disruption Events and Supernovae. Radio emission serves as a probe of the interaction between fast-moving ejecta and its surrounding. These observations provide key information (information that is rarely accessible in other wavelengths), such as the ejecta velocity, the density profile of the surrounding material, the nature of the ejecta (whether it’s a spherical or a relativistic jet), and can probe into the micro-physics shocks.
Tidal Disruption Events (TDEs) happen when a star travels too close to a supermassive black hole (SMBH). The tidal forces of the SMBH tear the star apart. Half of the material is being accreted back to the SMBH and half of it is ejected outside. This results in a multiwavelength emission. I study the radio emission from such events. Recently we found that the radio emission from some TDEs can rise years after the initial stellar disruption and the nature of this delayed radio flare is not yet understood. Possible explanations for such delayed radio emission are varying between delayed ejection of material, relativistic jets that initially point away from us, and a complex density structure for the close vicinity of the SMBH.
Selected publications on TDEs:
The peculiar behaviour of the first radio-bright TDE outside of a galactic nucleus
An off-axis relativistic jet emerges years after stellar disruption
A correlation between accretion and outflows in a TDE
At the center of stars, there are two mechanisms competing with each other - gravity and energy production by nuclear fusion of elements. When a massive star (at least 8 times the mass of our sun) can no longer maintain nuclear fusion in its core, gravity wins, and the core of the star collapses. This derives a blastwave in the surroundings of the progenitor star which radiates in all wavelengths. The radio emission provides key insights into the density profile of the environment and the mass-loss processes from the progenitor star that deposited it.
Selected publications on supernovae:
Statistical analysis of the first systematically monitored sample of core-collapse supernovae
Revealing a complex density structure around a Type Ib supernovae
In parallel to my observational work, I develop machine learning methods to extract physical information directly from time-domain data. My work focuses on classifying astronomical transients using only light curves, without relying on host-galaxy information or hand-engineered features. I built an LSTM-based classifier to distinguish tidal disruption events (TDEs) from Type Ia supernovae using ZTF alert photometry, demonstrating that the temporal evolution alone carries strong discriminative power. More recently, I have been exploring the use of large language models for scientific time-series analysis. In particular, I study how LLMs perform on transient classification tasks, how few-shot context affects their predictions, and what their reasoning reveals about the decision process. This work aims to bridge data-driven models with interpretable scientific reasoning.
See more in my blog posts on