Fraud Detection using Graph Machine learning (Jan 2022 – Feb 2022) (github)

  • A large financial dataset, which was the IEEE-CIS Fraud Detection Dataset provided by Vesta, was integrated, cleaned and analysed.
  • A graph machine learning end-to-end solution was developed to detect fraud
  • The dataset used was the IEEE-CIS Fraud Detection Dataset provided by Vesta on Kaggle
  • Two Relational graph convolution networks were developed and compared.

Monocular Depth Estimation using Transfer Learning with focus on complex scenarios (June 2021 – August 2021) (github)

  • Concept of Transfer Learning was used to develop three monocular depth estimation approaches, as an alternative to the currently used expensive and limited LIDAR solutions.
  • Three solutions (namely the Pix2pix model, the U-net with DenseNet encoder and the U-net with MobileNetv2 encoder) were produced and experimentally compared
  • The approaches were compared using a specifically engineered metric, in both complex indoor and outdoor situations, and the results were analysed.
  • A Graphical user interface was developed to facilitate the hybridised and appropriate use of the engineered models

Food and Nutrition Analytics (April 2022) (blog) (github)

  • Diet and nutrition data was analysed and visualized to derive insights
  • Dataset used was originally obtained from MyFitnessPal.
  • The insights were derived to aid supermarkets in better stocking of goods and to understand the diet trends and eating habits of users.

Query Optimizer (March 2021) (github)

  • A Query Optimizer and estimator were built for SJDB (A simple data base designed by Dr Nicholas Gibbins)
  • The Estimator accepted a database logical query and estimated the total cost in terms of disk accesses
  • The Optimizer optimised a given query by pushing down the Selects, creating Joins and adding Projects

HR Analytics (March 2020) (github)

  • HR dataset was analysed using R to predict if an employee would be promoted
  • Dataset used was the HR Analysis Case Study Dataset
  • The data was cleaned, analysed and visualized.

Tweet Classification on the MediaEval Benchmarking Initiative (January 2021)

  • The Twitter MediaEval data was analysed
  • Five tweet classification approaches were presented and critically analysed in depth
  • The approaches were then ranked based on this analysis

Data Swiss Knife (March 2020) (github) Team size: 3

  • Automated the data cleaning, modification, and model generation (machine learning models and deep learning)
  • The Deep learning models were automatically improved iteratively, based on the generated baseline model’s results

Wallee: Payment wallet (February 2019 – April 2019) (github) Team size: 6

  • Worked as the lead developer who assigned tasks to the other developers, developed backend modules and was responsible for seamless integration of the frontend with the backend
  • Django web framework and MySQL were used, and Scrum framework was followed for the development
  • Oversaw resolving of bugs that were discovered by the testers during the testing phase

HackMan (February 2019 – April 2019) (github) Team size: 3

  • A Hackathon Management System made to aid in better organization of hackathons
  • This website was built using Python and the Django web framework
  • It featured a participant dashboard, a judge dashboard and a cafeteria dashboard

UoS Book exchange (March 2021 – May 2021) (github) Team size: 2

  • An online book exchange made for the students of University of Southampton
  • This website was built using MERN stack
  • Functional requirements and architecture can be found in the github repo

Constructoo (March 2021 – April 2021) (github) Team size: 3

  • Constructoo is an Android application for construction deliveries
  • It assists the managers in a company by controlling the packages, determining the construction sites, and assigning the specified driver for this shipment
  • Functional requirements and architecture can be found in the github repo

Zoo Management System (January 2018 – March 2018) (github) Team size: 5

  • A Zoo Management Application built to aid in effective and efficient management of animals in the zoo
  • The application was built using Java Applets and the backend was made using MySQL
  • Separate portals for Manager, Staff, Head of zoo and the Vet were developed

HqOS: Threat Analysis (February 2021) Team size: 6

  • The product’s website and whitepaper were analysed, the vulnerable Assets involved were highlighted
  • Misuse cases and security cases were designed
  • System and Security requirements were elicited and DFD diagrams were used to illustrate the same
  • Finally, the top 10 privacy risks were picked with justified rationale and appropriate steps were suggested

Automated Negotiation Agent (January 2021) (github) Team size: 3

  • User and opponent modelling, along with bidding strategy were engineered for the agent
  • The agent was designed for a league with a closed setting and preference uncertainly enabled