Realtime Analysis for Twitter Data

March 2018 - Present

• The project aims at building a visual representation of how trending tweets spread over time.
• The visualization is intended to be done in real time.
• We also intend to use Machine Learning techniques to determine patterns as to why these tweets began to trend.

Neural Style Transfer

March 2018 - Present

• The process of using CNNs to migrate the semantic content of one image to different styles is referred to as Neural Style Transfer. The aim of the project is to build generative and descriptive models using VGG for generating image representations.
• In the descriptive model, the idea is to update pixels in the (yet unknown) stylized image iteratively through backpropagation, which starts from random noise.
• In generative model however, we first optimize a the model iteratively and produce the styled image through a single forward pass.
• The project aims to provide similar results to apps like Prisma, Ostagram, Deep Forger or Google's Deepdream.

Measuring a Cassandra Cluster

March 2018 - Present

• This project focuses on measurements across a Cassandra cluster for various topologies.
• We are particularly interested in measurements of latency and clock skew.
• The topologies consist of different arrangements of VMs, containers and physical hosts.

Contributions to GRDB

March 2018 - Present

• TBD

Comparative study of Statistical and Graph Based Recommendation Systems

August 2017 - December 2017

• The project is aimed at understanding various recommendation systems currently available - both machine learning based and graph based so as to identify use cases for each kind of system.
• We investigate statistical recommendation techniques based on collaborative filtering(user and item based). The addition of content and context is also evaluated through Bayesian models with algorithms like Probabilistic Matrix Factorization.
• We aim to fabricate a graph based recommendation system for movie data based on motivation from social networks.

Chat based Automated Sales Assistant

April 2015

• Used Python to design a chatbot to understand and answer natural language queries about mobile phones.
• The project involved building a Named Entity Recognizer using the Hidden Markov Model. A Relationship Entity Extractor was build using the Viterbi algorithm and finally a natural language response generator which converted the resultant query into natural text.

Framework for High Availability of Applications

April 2015

• Developed a framework on top of which applications could be written with assured availability and stateful restart so that it can run without any visible interruptions.
• Minimum downtime, heartbeating mechanism and stateful restarts were some of the key aspects of what the project handled.
• This framework was written in C and built in the Unix Environment. POSIX Compliance was assured as much as possible.

Hadoop on Openstack

April 2013

• Led a team of fifteen members on a project which involved autoscaling of Hadoop nodes in an Openstack environment based on the number of pending Hadoop map and reduce tasks

Regex Comparison Engine

April 2013

• This tool aimed to compare two regular expressions and decide if they generate the same language. It took two regular expressions as input, built their respective parse trees, built the non-deterministic finite automata, converted them to deterministic finite automata, minimised and checked them for equivalency as the final step. My role specifically involved building the parse trees and the NFAs.

Green Cloud

August 2014

The objective of the project was to minimise the energy consumed by the Swift architecture by controlling the writes into the replicas that are present for fault tolerance.

Visual Tweet Analysis for Trending Twitter Data

March 2018 - Present

• The project aims at building a visual representation of how trending tweets spread over time.
• The visualization is intended to be done in real time.
• We also intend to use Machine Learning techniques to determine patterns as to why these tweets began to trend.

Neural Style Transfer

March 2018 - Present

• The process of using CNNs to migrate the semantic content of one image to different styles is referred to as Neural Style Transfer. The aim of the project is to build generative and descriptive models using VGG for generating image representations.
• In the descriptive model, the idea is to update pixels in the (yet unknown) stylized image iteratively through backpropagation, which starts from random noise.
• In generative model however, we first optimize a the model iteratively and produce the styled image through a single forward pass.
• The project aims to provide similar results to apps like Prisma, Ostagram, Deep Forger or Google's Deepdream.

Measuring a Cassandra Cluster

March 2018 - Present

• This project focuses on measurements across a Cassandra cluster for various topologies.
>
• We are particularly interested in measurements of latency and clock skew.
• The topologies consist of different arrangements of VMs, containers and physical hosts.