Sriharsh Bhyravajjula

Hi, I’m Harsh. I wear many hats.

Hat One. I’m a Senior Applied Scientist at Outreach, where I help build and evaluate LLMs and agentic systems for sales meeting intelligence. Before this, I completed my Master’s in Computational Linguistics at the University of Washington, Seattle. Before that, I worked as an NLP Data Scientist at Klevu, building search, discovery, and analytics for ecommerce. And even before that, I studied Computer Science at IIIT-Hyderabad, India, where I developed a strong foundation in machine learning and natural language processing. I have almost six years of experience in building practical AI systems that demonstrably solve real-world problems, with the last couple of them focused on large language models and generative AI.

Hat Two. I’m an affiliated researcher at Antoniak Lab at the University of Colorado, Boulder. I am deeply interested in computational poetics and narrative understanding — exploring how computational methods can help us understand literature, poetry, and storytelling. Recently, I worked on analyzing whitespace in poetry and its implications for LLMs (EMNLP 2025). I was previously inspired by Kurt Vonnegut’s talk on the Shape of Stories and developed MARCUS, a system for extracting character arcs from narratives, which won the Best Paper Award at Text2Story@ECIR 2022.

Hat Three (this might just be a bad haircut, the shtick gets old after three hats anyway). I write poetry, some of it has been published (I think that makes me a poet? I’m not quite sure and no one seems to really know. So sure, I’m a poet). I spend a lot of time at Hugo House talking to people about poems and storytelling. My poems have appeared in Indian Literature, Madras Courier, and Muse India. My first poetry collection, sorting through paperwork after midnight was published by Authorspress, 2019. I apologize if you were one of the unsuspecting readers.