Machine Learning and Big Data

Music Recommendation System, Machine Learning

Collaborated in a team of 4 to build a Music Recommendation System, a KDD Cup 2011 Challenge to predict music ratings and identify favourite songs.

The datasets provided: Yahoo Music Dataset(1999-2010) were first used to standardize and pre-process the data based on Feature Scaling. Implemented Collaborative filtering(KNN), Content Based(Neural Net) and Ensemble methods(Adaboost) by calculating the accuracy metrics like RMSE(Root Mean Square Error) and classification metrics like Precision, Recall, F-measure.

In conclusion, Collaborative filtering- KNN proved to be the best machine learning algorithm for recommendation system with 95% accuracy and 21% RMSE.

Artificial Intelligence and Deep Learning

AI enabled Smart Sudoku Solver, Artificial Intelligence

The purpose has been to develop more effective algorithm in order to reduce the computing time and utilize lower memory space.

Implemented Sudoku solver using effective algorithms: Constraint satisfaction problem and Backtracking search. Overhauled performance by using forward checking and heuristics reducing time and space complexity.

The programming language and forms used are C#. NET programming language. C# is used for developing business logic and user forms have been creating to develop GUI and interface for better understanding of sudoku solver. Xamarin Studio has been used for implementation, which is a Visual Studio alternative for Mac.

Cloud Computing and AWS

Developed a serverless application and portfolio on AWS using S3, DynamoDB and Lambda.

Web Development

Library Management System(Web Based), Database Design

Developed a fully functional web based Library management system. Developed Web Services which are invoked via API calls to Check-ins, Check-outs, inventory, borrowers and late fee management. (Hapi, joi, boom, mysql modules)

Designed a database involving formulation of ER/EER models. Developed Scripts to Clean Data and load to Mysql using (Pandas, Pymysql, numpy modules)

Virtual Reality and Computer Vision

Vision Based Heartbeat Estimation, Multimedia Systems

Collaborated in a team of 2 to build an intuitive application that estimates heartbeat of a user using high resolution webcam of a laptop.

In this project, we have used PhotoPlethysmoGraphy technology along with Independent Component Analysis to determine the heart-rate of an individual using their own webcam in real-time. This application uses the OpenCV library to capture and find the location of the user's face, and then to isolate the forehead region. Data is collected from this location and analysed over time to estimate the user's heart rate. This is done by measuring average optical intensity in the forehead location, in the sub-image's RGB channel. With good lighting and minimal noise due to motion, a stable heartbeat can be isolated in about 30 seconds.

Database Design and Operating Systems

Database Engine, Database Design

The goal of this project was to implement a (very) rudimentary database engine that is loosely based on a hybrid between MySQL and SQLite, "DavisBase".

Implementated using Java and MySQL and operated entirely from the command line and API calls (no GUI). The database support actions on a single table at a time, no joins or nested queries. It used file-per-table approach to physical storage. Each database table will be physcially stored as a separate file.