Khushi Agrawal

About Me

I like researching the internals of machine learning frameworks, compilers, parallel computing, and deep learning. I'm fascinated by how these fields connect, and I'm always exploring how they can be used to build more efficient systems.

I started contributing to PyTorch out of curiosity and obsession with understanding how things work under the hood. Here's the link to all my pull requests. Currently, I'm working as an Open Source Machine Learning Engineer on the PyTorch Team at Hugging Face. Previously, I worked at Mercor as a Software Engineer and before that I worked as a Compiler Developer with Ondřej Čertík on adding new features to the LPython and LFotran compilers. I've also contributed to the CuPy (via GSoC and Open-Source), where I started an initiative to add interpolate functions in CuPy and worked on various experimental fusion and kernel methods. Before that, I interned at Quansight-Labs, contributing to the NumPy and discovering multiple compilers and interpreters.

In my free time, I enjoy drawing and writing, and I mostly love travelling, exploring different places, and playing adventurous games.

Hugging Face

Open Source Machine Learning Engineer; PyTorch Team

Oct'26 - Present

Mercor

Software Engineer

Jun'24 - Sept'26

GSI Technology

Compiler Developer; Contributed to LFortran and LPython

Sept'23 - March'24

Google Summer of Code

Student Software Developer; Contributed to CuPy

June'22 - Sept'22

Quansight Labs

Software Engineer Internship; Contributed to NumPy

July'21 - Sept'21