Hi!

I am a final year undergraduate student at Visvesvaraya National Institute of Technology, pursuing a degree in Electrical and Electronics Engineering. During my junior year, I had the opportunity of working with Prof. Konda Reddy at IIT Guwahati on Coreset Extraction techniques.

I am also an active member of IvLabs, the AI and Robotics Lab at VNIT led by Prof. Shital Chiddarwar. IvLabs has played a significant role in nurturing my research interests, where I worked on various projects, mentored juniors, and organized IEEE workshops on robotics. I also had the privilege of serving as the Vice-Chairman of our lab.

I am passionate about continuously improving my skills and expanding my knowledge. During my free time, I enjoy exploring my interests in programming and automation, contributing to open-source projects such as TorchVision, and creating Discord bots.

When not immersed in academics, I enjoy watching anime, playing basketball, and taking pictures of nature.

Email  /  CV  /  Google Scholar  /  Github  /  LinkedIn

I'm interested in the areas of Unsupervised Deep Learning, Computer Vision, Domain Adaptation and Robotics. I am open to collaborations and looking forward to research opportunities in similar domains, and I am also willing to explore new fields such as Natural Language Processing and Graph Neural Networks. Feel free to contact me.

This work introduced a new setting for unsupervised domain adaptation and utilized a prototypical network to create domain embeddings. We used the Local Outlier Factor (LOF) to pseudo-label unknown classes, and used graph neural networks with attentional aggregation for the adaptation stage.

This project implemented Triplet Network and FaceNet algorithms from scratch with ResNet as the backbone architecture for face recognition. It was developed at IvLabs for one-shot and few-shot learning on different datasets and with the goal of deploying it as a face recognition-based door lock system.

This project is an implementation of a sliding window technique with a two-stage detector, enhanced with the Overfeat framework, for object detection on a Raccoon Dataset.

This project involved implementing a PID controller on a TurtleBot robot to perform a Goal-to-Goal task and Path Planning task. The controller was modeled using a Hermite curve to generate a trajectory for the robot's motion from the initial to the goal pose.

Ofcourse, I haven't build this website from scratch. This is a courtesy of Jon Barron.