Dhruv Sondhi

I am pursuing a Master of Computer Science with interdisciplinary research in astrophysics and machine learning, focusing on automated identification and classification of astrophysical named entities using large language models.

I integrated the Universal Astronomy Thesaurus (UAT) to enrich an ADS dataset, developed data-augmentation techniques including duplication and substitution of context-important entities, and fine-tuned models such as AstroBERT to improve vocabulary coverage and downstream performance.

During a Mitacs Global Research Internship I implemented NLP on NASA ADS literature to identify trends across astrophysics subdomains and assessed machine learning usage; I developed ML models and researched text-processing techniques including RNNs and deep-learning approaches to retrieve relationships from large corpora.

As a Google Summer of Code contributor to the TARDIS scientific Python project I developed logging and packet-tracking frameworks, used Numba for JIT performance, and contributed features to the main project branch; I have presented work at NLP for Space (ESAC), ADASS 2024, CASCA Toronto, and OGMC-UWaterloo.