Hello! I am Varun, a PhD student at New York University's Center for Data Science,
advised
by Prof. Eunsol Choi and
Prof. He He.
I spent two years at Google DeepMind, India as a pre-doctoral researcher working on
making
LLMs better and
faster. I was fortunate
to work with brilliant researchers in the Machine
Learning and Optimizations team led by Dr.
Prateek Jain. During my time there,
I worked on extracting unsupervised feedback
from text to improve LLMs at various tasks with Time-Reversed Models, advised by
Dr.
Karthikeyan Shanmugam
and Dr. Arun
Suggala.
In addition, I have worked on optimizing the attention layers for million-context
models to enable efficient inference. I also developed methods to accelerate the FFN and SoftMax
layers under the guidance of
Dr. Praneeth Netrapalli.
Before that, I collaborated with
Prof. Vivek Gupta on supplementing LLMs with
knowledge graphs and testing the reasoning abilities of Visual LLMs.
I also spent a summer at Amazon
Science, modeling customer demographics through their behavioral
history.
I developed my interests in Machine learning and its applications during my undergrad at
IIT Guwahati, where I participated in many hackathons and challenges.
Along with numerous kaggle wins, my team
also placed second
in Amazon Machine learning challenge
2021.
News and Updates
2025
August
Starting my PhD at New York University!
Apr
Attending ICLR 2025 in Singapore! See you there!
2024
Dec
Attending NeurIPS 2024 in Vancouver, Canada!
Oct
Presenting Time-Reversed LLMs, an awesome work with Karthikeyan,
Arun, Rahul and Sravanti
Won the spotlight award at NeurIPS 2024! 🎉
May
Joined Amazon Science as an Applied Scientist Intern at Bangalore!
Won the best paper award at DeeLIO Workshop @ ACL 2022 for my work with Vivek Gupta!
April
Published my work with Hiroyuki Takeshita and Yuji Iwahori in Journal of Imaging, MDPI
My Team won Silver Medal at Inter IIT Tech Meet 10.0
organized
by
IIT Kharagpur!
Oct Conducted a workshop
on
the
basics of Transformer architecture at MLTechFest'21 organized by Tensorflow User Group
(Mysuru)
(Meetup!) &
Google Developer's Group (Mysuru) (Website).
Sep Conducted a Workshop session on
Computer
vision - OpenCV basics - by IEEE SFIT & IEEE APSIT.
