Hi, I’m Raj, a CS PhD student at Cornell Tech. I am advised by Emma Pierson and supported by an NSF graduate fellowship. I’m broadly interested in the social implications of machine learning, usually focusing on empirical questions in NLP and/or healthcare. My recent projects study how language modeling research is becoming increasingly sociotechnical and how coarse race categories undermine algorithmic fairness audits.
Previously, I studied CS at MIT with minors in Women’s & Gender Studies and Biology. I was lucky to work with the Data + Feminism Lab, where we developed intersectional, participatory NLP systems to support activist labor (FAccT 2022). I interned with Apple’s Siri Research team, studying efficiency and compression for language models [1, 2]. I am very thankful to my excellent mentors along the way, including Harini Suresh, Jonathan Frankle, and Shayne Longpre; see here for a longer list of people who I’m indebted to! I was also president of AI@MIT, where I helped design intro ML workshops and ordered lots of food.
I hew closely to PhD stereotypes: my primary hobbies are baking, rock climbing, reading, mixology, and tennis. If you’d like to read some unreliable recipes I’ve written, see here.
Updates
- [2023-11] Data (for 17K LLM papers through Sep 2023) and code for the LLM survey paper are now available on Github! I’ll be presenting a poster at TADA 2023.
- [2023-08] I presented our paper on granular race disparities at MLHC 2023! Thanks for the excellent conversations; code is available on Github.
- [2023-07] New working paper on arXiv! We analyzed 14K language modeling-related papers on arXiv to detail recent publishing trends.
- [2023-04] Our new paper is on arXiv, the first from my PhD! We find that granular race categories are critical to algorithmic fairness analyses in healthcare; here’s a summary.
- [2022-08] Our paper on compressing language models is accepted to COLING 2022! I led this project during an Apple internship in 2021.