Anirudh Mukundan Raghavan

Software Engineer

About Me

I am an aspiring Software Engineer, with a passion for Machine Learning and Web Development.

Graduate Research Assistant,Machine Learning + Software Engineering, Feb 2020-Present

Worked single-handedly on Human Activity Recognition using Deep Learning for ICU Patient The aim of the project was to recognize human poses based on video frames collected from anRGB and depth camera placed in the ICU . Responsibilities included designing neural networkarchitectures to recognize poses and customizing existing libraries such as YOLO, OpenPose andDensePose for our dataset. Achieved a precision of 0.91, recall of 0.85 and an F-1 score of 0.8 Most of the code was written in Python, involving use of libraries such as PyTorch, Tensorflow and Pandas.

Wrote and maintained a custom Java GUI annotation application to annotate video frames received from the cameras. Used by a team of annotators in the Intelligent ICU Team. Written in Java, using JavaFX.

Worked as a database administrator and was responsible for designing a MongoDB database schema for EHR(Electronic Health Record) data generated in the ICU. Wrote scripts in Python to automate transfer of data to the database, as well as generate weekly statistic reports for the database.

Capstone project for Bachelors degree.Worked in a team of two to help build a Deep Learning solution to identify landmarks from images. Used Squeeze and Excitation networks in conjunction with Residual Neural Networks on Google Landmarks data set and achieved an accuracy of 92%. Built a user interface for the system using Flask. Written in Python, using the Keras framework.

A web application which aggregates news articles from multiple sources and places them in one convenient location. Supports features such as user registration, user account customization(selecting news sources) and automatic deletion of articles once they are read. Built in Python using the Django framework, using PostgreSQL as the backend database and AWS S3 for static file storage.

Implemented consistent hashing, peer-to-peer communication algorithms(Gossip and Push-Sum) and a Distrbuted Hash Table using the Tapestry overlay network.

A distributed engine for a twitter-like application with Phoenix and Elixir and a simulations of its performance with varying number of concurrent connections using concurrent actors as independent nodes. Supported features such as tweeting, retweeting and live querying of tweets. Written in Elixir, using the Phoenix framework.