Success consists of going from failure to failure without the loss of enthusiasm.  - Winston Churchill

Maghav Kumar

I am a recent MS graduate from UIUC Computer Science. In my time at UIUC, I mainly worked with Professor Alex Schwing. My research interests lie at the intersection of Computer Vision, Generative Models and Reinforcement Learning. Recently I've been more interested in Generative Models for Zero-Shot and Few-Shot scenarios. I spent the Summer of 2018 at Baidu Research, Sunnyvale.

I am also excited by recent Machine Learning startups and their impact on real world problems. I spent time at Planet Labs and Hive AI. Currenly, I'm working at C3.ai.

Before working on Machine Learning and Computer Vision, I worked on C++ Language Development.

E-Mail | Github | LinkedIn

(I'm not good with pictures)
Publications
sym



[NEW] A Singular Value Perspective on Model Robustness
Malhar Jere, Maghav Kumar, Farinaz Koushanfar
Under Review
pdf
sym

Can We Learn Heuristics For Graphical Model Inference Using Reinforcement Learning?
Safa Messaoud, Maghav Kumar, Alexander Schwing
Computer Vision and Pattern Recognition (CVPR), 2020
(Oral presentation)
pdf
Will also present at CVPR'20 Deep Vision Workshop.

News
Dec 2020 Joined C3.ai for building and helping out with the future of Enterprise AI.
Nov 2020 Posted our work with friends from UCSD to arxiv about understanding the gap between human vision and CNNs using SVD.
March 2020 Paper accepted at CVPR 2020 as oral presentation.
Dec 2019 Joined Hive AI for using Machine Learning to defend online platforms from abuse and hate speech prevention.
June 2019 Our Team Schwail came 2nd in the TextVQA Challenge hosted at the VQA Workshop at CVPR 2019.
May 2019 Starting Internship at Planet Labs to work on global semantic segmentation.
Aug 2018 TA'ing and helping to create the Deep Learning course at UIUC under Prof. Lazebnik.
May 2018 Spending the Summer at Baidu Research, Sunnyvale to work on Neural Architecture Search.
Teaching

Some great people I was fortunate to be mentored by: Robert, Greg, Yanqi, Vincent.

Some great people I was fortunate to work with: Daniel and Jyoti.

Template credits: Jon, Saurabh, Unnat and Jyoti.