Yash Vanjani

I am a Machine Learning Engineer at Meta Superintelligence Labs. Previously, I was at Essential AI where I scaled foundation model pre-training across thousands of AMD GPUs and Google TPUs. My work there spanned distributed training optimization, inference systems, and data curation, contributing to several publications on efficient LLM training and optimizer innovation.

Before that, I worked at Nutanix on Distributed Storage Systems (Core Data Path - Stargate), and at Adobe on Adobe FrameMaker.

I hold a Master of Science in Machine Learning & Systems from Carnegie Mellon University and a Bachelor of Technology in Computer Science & Mathematics from IIT Guwahati.

I am broadly interested in making LLM training more efficient — from optimizing distributed infrastructure to understanding why certain architectural choices work, and using that intuition to design and scale the next generation of models.