I am a third year PhD student in Electrical and Computer Engineering, Purdue University; currently working under Prof. Eugenio Culurciello as part of e-Lab. Most of my research hours are invested in developing and training CNN/RNN architectures in supervised, unsupervised, as well as semi-supervised way.
I received my undergraduate degree in Electronics and Communication Engineering from Indian Institute of Technology (IIT) Guwahati, India in 2012. During my undergraduation I focused mainly on robotics. I was the team leader of IIT Guwahati ROBOCON 2011 team. In 2011 I went to Hanyang University for my summer internship which was funded by them. Based on the work done by me, I was offered to join Hanyang University as Masters student on scholarship. As a result I went to South Korea and received my Masters degree in 2014. While working on robotics, I became more inclined towards the vision part of it. Pursuing my interest in this direction, my Masters thesis was on Fuzzy Logic and Systems.
Currently I am mostly interested in developing deep neural network architectures which not only stand out in terms of accuracy metrics, but can also be used in real life applications. Scene understanding is an important part of any machine learning task and I am working on this on two fronts: using semantic segmentation or end-to-end learning. Some of its applications can be found in self-driving vehicles, augmented reality, etc. This serves as a major motivation for me to develop networks which can not only be run on GPUs but also on embedded devices such as NVIDIA TX1.