Saurav Joshi

I am currently a 2nd year Masters student specializing in Data Science at the University of Southern California. An essential part of my academic journey is my involvement in research at the USC Information Sciences Institute where I am advised by Prof. Filip Ilievski and Prof. Jay Pujara. Here, I delve deep into the realms of knowledge graphs, natural language processing, and the vast expanse of social media. My research pursuits do not stop at just this. I’m engaged in a collaborative DARPA project, a joint endeavor that spans the expertise at USC ISI and ICT. The crux of this project revolves around knowledge management and granular knowledge retrieval. Within the broader spectrum of NLP, my research aspirations center around developing explainable and efficient agents tailored for a range of tasks. Simultaneously, my affinity for cloud technologies drives me to integrate them for enhanced scalability and speed in various projects.

Parallel to my academic responsibilities, I’ve been actively participating in the Google Summer of Code. In its 2023 edition, I embarked on a project titled “Question-Answering over DBpedia with Pretrained Auto-regressive Models”. The primary aim was the continued pretraining, finetuning, and optimization of an opensource code model, notably known as Starcoder1B. This engagement followed my previous successful stint in the 2022 Google Summer of Code, where I contributed to “Template Discovery for Neural Question Answering over DBpedia”. This ambitious project had set its sights on crafting a diverse dataset, one that boasted millions of rows. The ultimate goal was to ensure the efficient translation of natural language questions into SPARQL.