Welcome to my website!

This website is designed to showcase my skills, interests and projects

Born in Chennai, India, I completed my undergraduate education majoring in Electronics and Communications Engineering from Anna University. I'm currently pursuing my master's degree in Computer Science from Virginia Tech, USA. I'm interested in software development, distributed systems, Machine learning and Computer Vision.

>

Things I Can Do

I have listed a few of my abilities which makes me employable for my skill

  • Programming (Java, Python, C/C++, JavaScript)
  • distributed systems
  • Machine learning
  • Computer Vision

A Few Accomplishments

A few of my recent research works and projects listed below,

Distributed storage system

Built an extensible serverless storage system for the serverless edge computing system which eliminates the need for an external storage system for storing the functions’ data-objects.

Social Distance detector

Built a Social distance detector that automatically identifies social distance violations based on a video from a monocular camera. Used perspective projection to obtain the Bird’s eye view of the population and calculated the distances based on the centroid from the bounding boxes provided by the object detection YOLO v3

Multi-modal search engine

Developed a search engine application, which uses Okapi BM25 model to rank the documents based on the query statement. Created the backend API running on the BM25 model using Flask and integrated to the frontend UI built on react

Further directions

There are a few areas that I'm working on and would love to learn more about. They are mentioned in this section below,

Distributed consensus algorithms

Studying and researching more on distributed consensus algorithms like paxos and raft. Looking for ways to bring consistency and availability in the same system.

Federated ML

Learning about Federated ML and working on relevant implementation projects on Federated ML.

Serverless MapReduce

Trying to implement MapReduce on a serverless system. Overcoming challenges such as the small memory and computing of the lambda, battling timeout limits and the maximum concurrency limits in lambda.

Semi-supervised learning

As data grows in the real world, the cost of annotating or labelling data also increases. Working on semi-supervised learning to overcome this drawback of expecting expensive labelled data and use the cheaply available unlabelled data to train our model.