My research focuses on resource efficient deep learning. Specifically, I try to find ways to make deep learning models amenable to resource-constrained scenarios (both labeled/unlabeled data and compute). In the past, I have developed methods pertaining to synthetic data augmentation and self-supervised learning for effective downstream learning in low-resource (labeled data) scenarios. Currently, I am also looking at scaling laws of using sythetic data to train SLP models including effective evaluation of sythetic data. My papers are written around applying resource efficient methods to the diverse tasks in the broader domain of Speech, Language and Audio Processing (SLP), including but not limited to the tasks of NLU, RIR estimation, audio generation, compositional reasoning, audio captioning, etc.

I am always open to collaborations, and please feel free to drop me a mail!

Google Scholar Semantic Scholar

Pre-prints

Audio and Spoken Language Processing (Chronological)

Natural Language Processing (Chronological)

Workshop