Zain Merchant

I just graduated from the University of Southern California with a B.S. in Computer Science.

Previously I was on the ML Infra team at LinkedIn, where I worked on training LLMs and the Liger Kernel. Before that, I worked with Jason Calacanis and the LAUNCH team on scalable audio indexing, and worked at several startups on machine learning teams.

I'm currently doing research at the GLAMOR lab at USC and am broadly interested in building performant infrastructure, LLMs, and software for robots.

Email  /  LinkedIn  /  Twitter  /  Github  /  Google Scholar

profile photo

research
project image Efficient Evaluation of Multi-Task Robot Policies With Active Experiment Selection
Abrar Anwar, Rohan Gupta, Zain Merchant, Sayan Ghosh, Willie Neiswanger, Jesse Thomason
Evaluating robot control policies across many tasks is time-consuming and costly. We propose framing this as an active testing problem, using natural language to model task similarities and guide efficient experiment selection. Our cost-aware framework reduces evaluation effort by prioritizing informative trials, supporting both continuous and discrete outcomes.

Preprint, 2025
arxiv twitter
project image Generating Contextually-Relevant Navigation Instructions for Blind and Low Vision People
Zain Merchant, Abrar Anwar, Emily Wang, Souti Chattopadhyay, Jesse Thomason
Navigating unfamiliar environments presents significant challenges for blind and low-vision (BLV) individuals. We investigate how grounded instruction generation methods can provide contextually-relevant navigational guidance to BLV users.

ROMAN 2024 Interactive AI Workshop (Best paper), 2024
paper