Under the supervision of Dr. Faysal Hossain Shezan, I am working to develop a tool that would perform symbolic execution on Tensorflow code. The approach generates a test harness for testing particular tensorflow APIs and extracts its related dependencies.
View on GitHubI developed a web-based tool that can evaluate boundary point segmentation algorithms on a LiDAR point cloud dataset of building roof structures. The tool I implemented evaluates three different boundary point segmentation algorithms: Neighborhood Based Approach, Concave Hull Algorithm and Delaunay Triangulation Based Algorithm.
View on GitHubWe developed a Convolutional Neural Network based approach to detect crop diseases from images of plant leaves.
Read MoreI developed the backend of the BICM (Bangladesh Institute of Capital Market) website which displays real-time and historical stock exchange data with visual insights. My role involved implementing API endpoints to scrape and process stock market data from the Dhaka Stock Exchange (DSE) website. After development, I conducted a thorough vulnerability analysis of the project, discovering critical vulnerabilities. For instance, improper exception handling in file operations exposed file paths in case of an error. Additionally, missing data validation steps caused the server to process unexpected inputs when the DSE website layout changed. The vulnerabilities could cause the website to crash and expose sensitive information from an important government website. I addressed the issues by adding validation and exception handling steps.
Check WebsiteI developed a party planning web application which serves as a centralized hub, connecting various party service providers (caterers, decorators, vendors etc.) with potential customers. Additionally, we developed a recommendation system using K-Nearest Neighbor algorithm to suggest a complete plan for services to avail based on user inputs like date, budget, and location.
View on Github