Hello! My name is Saurabh Garg and I am a second year PhD student at Machine Learning Department at CMU where I am advised by Prof. Zachary Lipton and Prof. Siva Balakrishnan. I am broadly interested in building robust and interpretable machine learning systems. Machine learning algorithms are typically developed and evaluated under simplistic assumptions that are often violated in practice. I am interested in understanding the behavior of machine learning models in real-world scenarios and building provable methods to make progress towards relaxing simplifying assumptions in order to make robust and trustworthy models.

I did my undergrad from IIT Bombay, India with major and honors in CS and minors in Applied Statitics in 2018. After that, I spent one amazing year at Samsung Headquaters, Korea. In the past, I have worked with Prof. Suyash Awate on building statistical machine learning algorithms for exact MCMC samspling as a part of my Bachelor’s thesis. During my stay at IITB, I have also spent major time working with Prof. Preethi Jyothi on the problem of building robust language models for code switched speech. I was also fortunate to work with Prof. Soumen Chakrabarti on building interpretable question answering systems using KG and corpus.


Aug 2021: Continuing my colloboration with Hanie Sedghi and Behnam Neyshabur at Google Brain as a Student Researcher.
June 2021: New work on learning from positive and unlabeled data will be out on arxiv soon.
May 2021: Two papers at ICML: (i) Work on obtaining generalization bound with unlabeled data got accepted as Long talk at ICML 2021 [Paper]; (ii) Work on understanding heavy tails in PPO also to appear at ICML 2021 [Paper].
April 2021: Our work on obtaining generalization gaurantees with unlabeled data will be presented at RobustML Workshop at ICLR 2021 [Paper] [Poster].
April 2021: Our work on understanding behaviour of gradients in PPO will be presented at SEDL Workshop at ICLR 2021. [Paper] [Talk] [Poster].
Feb 2021: Excited to be interning with Hanie Sedghi and Behnam Neyshabur at Google Brain during Summer 21.
Feb 2021: New work on understanding behaviour of gradients in PPO is out on arxiv.
Sept 2020: Our work on label shift got accepted at NeurIPs 2020 [Paper] [Poster].
July 2020: Our work on label shift estimation was accepted as Oral at ICML UDL 2020 [Talk] [Full Paper].
April 2020: Our work on Neural Architecture for Question Answering was an invited Oral at ECIR 2020 [Talk].
June 2019: I will be joining CMU ML Ph.D. in fall 2019.
April 2019: My B.Tech thesis titled "Estimating Uncertainty in MRF-based Image Segmentation: An Exact-MCMC Approach" got accepted at Medical Image Analysis 2019 journal
Dec. 2018: Received Excellence in Research Award from CSE dept, IIT Bombay
Nov. 2018: Presented my paper"Code-Switched Language models using Dual RNNs and Same-Source Pretraining" at EMNLP 2018, Brussels (poster)
Oct. 2018: Paper titled "Neural Architecture for Question Answering Using a Knowledge Graph and Web Corpus" got accepted at Information Retrieval Journal
Sept. 2018: Moved to Suwon, South Korea and joined Samsung Research Korea as Engineer
Sept. 2018: Presented my paper "Dual Language Models for Code Mixed Speech Recognition" at Interspeech 2018, Hyderabad (poster)
Aug. 2018: Graduated from IIT Bombay.
May 2018: Paper titled "Uncertainty Estimation in Segmentation with Perfect MCMC Sampling in Bayesian MRFs" got accepted at MICCAI, 2018 (poster)
Dec 2018: Invited to spend two weeks at Microsoft Research India to work on Indian language technologies with Prof. Preethi Jyothi
May 2017: Internship @ Samsung Research Korea
May 2016: Internship at Purdue Univeristy, US advised by Prof. Alex Pothen
July 2015: Changed branch from Electrical Engineering to Computer Science Engineering
July 2014: Joined IIT Bombay