I believe that with the amount of data available (and constantly generated) today, one can make a positive impact to better the lives around us. Aspiring to be a good Data Scientist and a better person everyday.
I hold a Bachelor’s (Computer Science) and Master’s degree (Data Science) and a 3+ years of track record of analysing business intelligence reports and data that are extracted from different sources in the retail domain with ability to make data driven decisions and convey insights via compelling storytelling presentations. Proven knowledge in expanding existing data collection and data delivery platforms. Worked extensively in development report and maintenance of analytical projects using Python, Java, SQL and PowerBI tools.
Awarded Masters degree with Distinction
Completed bachelor of engineering degree in computer science under the faculty of information and communication engineering and placed in First class
This project was undertaken to do stance classification of tweet and discovering the best fitting deep learning model that can successfully classify stance (FAVOR, AGAINST, NONE) based on particular Target (topic) about a tweet.Final selection was made of LSTM model as they proved to work best in natural language processing scenario. Analysing intra variability of classes and imbalancing of data, efforts were taken to choose a balanced data set for training purpose. glove MODEL trained on wikipedia data was used .The data was transformed to task vocabulary to fit the model. The model after large iterations of fine tuning made it robust and provided an accuracy score on validation as 52.97 and test score as 53.8 respectively.View Project
This project was undertaken to discover the best fitting machine learning model which can successfully classify images based on two factors, namely shape and labels. Three different ML model development techniques were employed : MLP, CNN . The CNN model proved to be highly accurate .Analysis of intra variability of classes and comparison of classification performance of four Convolutional Neural Networks based on parameter tuning gave us the optimal model. The images provided are grey scaled and sized to a 28*28 density. The development of the model has three stages: image preprocessing, detection, and recognition. The model demonstrated a promising result with an accuracy of 81 %for shape classification and 62% for label classification. Independent evaluation was done by collecting images from internet sources and real time clicks of victorian, Melbourne streets.View Project
95% of Australian sugar is grown in Queensland, and sugarcane is the second-highest export crop after wheat as it contributes 2.5 billion dollars to the Australian economy. Developed the time series model (FB Prophet) to forecast the overestimation and underestimation of sugarcane for sugar mills using Satellite imagery datasets of sugarcane fields at Queensland. Processed the image data using python libraries and Google cloud services and also merged that dataset with weather and soil quality datasets to find the correlation and improve the performance of the model.View Project
Developed an application using Amazon web services and extends a Web system to add business functionality. The environmental setup was done using elastic beanstalk where the backend was developed using NodeJs. The virtual server was launched through the EC2 instance and NoSQL database mongo DB was used for data storage. A complete user profile authentication was set up using amazon cognito.View Project
Developed a housing rental system using Java programming language with the help of Eclipse IDE. Also to make records accessible via Java JDBC technology. This is a simple yet complete hotel management system. This system performs all the necessary tasks that a hotel software application performs. JavaFx was used to tailor make the Graphical User Interface. Text file handling was implemented in the program to import and export data to facilitate analysis.View Project