Projects
Below I have shared some of the projects that I have done over my professional and academic experience.
Blockchain and NoSQL
Student/Management
Mentored and led a team of 10 students in a research/start-up project to develop a platform to match customers and service providers using blockchain and artificial intelligence using MongoDB and Java.Developed a Blockchain using NoSQL/MongoDB technologies on cloud following the PBFT consensus mechanism .
Worked from scratch on how to develop successful and secure multi-link chains using Cloud Atlas MongoDB and Java. Currently the project is its nascent stages and conversion to a full fledged startup is high possibility.
Student
Worked on project in a diverse team with 5 students to analyse the sentiments of comments and reviews on a ecommerce website for a NGO Petfinder.my which helps people adopt abandoned animals. We wanted to understand what are the prime reasons that a animal was adopted over another animal. We used basic statistical techniques for the quantitative data and used VADER sentiment analysis package to get more value out of the text data. We then used these numbers to predict adoption rate for a animal.
Twitter Sentiment and Economic Analysis
Self
An ongoing project in which we are trying to predict the Dow Jones Index Average using the mood of people by looking at twitter data on any given day. For this we automated the extraction of tweets and DJIA from Twitter and Yahoo finance, and used AWS for storing them. We gathered around 186,000 tweets and are currently using deep learning models to estimate the moods of the people. We want to understand how the mood of the nation affects the economic indices of a country.
Student
Classified a fraudulent firm by analyzing data using multiple ML techniques on 777 firms from the Auditor Office of India using Python with a 92% accuracy. Used multiple machine learning techniques such as Regression for forecasting the risk measures about a firm and understand the different factors that increase the risk factor for a company to be fraudulent. We then also used Classification techniques to get which company is Fraudulent and which one isn't.
Share of Wallet Model
Decision Scientist/Homedepot client
Designed a predictive model by using a combination of K-prototype clustering and quantile regression to estimate Share of Wallet for client to effectively target high opportunity customers
