I developed Algo-sphere, an interactive platform for learning data structures and algorithms through visualization, step-by-step explanations, and AI-powered debugging. The platform helps users understand complex algorithms with real-time visualizations and guided problem-solving.

It includes algorithm exploration, interactive execution breakdowns, and an AI debugger for code optimization. Built with React.js, Node.js, and OpenAI GPT, it provides a gamified learning experience with achievements and leaderboards.

This project showcases the application of AI in education, making algorithm learning more intuitive and accessible.

Link: github.com/zorogotty14/Algo-sphere

I developed an AI-powered tennis match analysis application using OpenCV and deep learning models to detect courts, track players, and extract gameplay insights. Designed for sports analysts and researchers, the system provides real-time data visualization and analysis to enhance match evaluation.

The application integrates YOLOv8, TrackNet CNN, and ResNet-18 for player tracking, ball detection, and action classification. Additionally, Azure OpenAI Services are used for advanced analytics, while deployment is managed through Azure Virtual Machines.

This project showcases the application of computer vision and AI in sports analytics, offering real-time match insights and automated performance evaluation.

Link: github.com/zorogotty14/CourtMetrics

I implemented an AI-driven Snake Game using Deep Reinforcement Learning, where an agent learns optimal strategies for movement and survival. The model leverages Deep Q-Networks (DQN) and reinforcement learning techniques to make intelligent decisions based on game state inputs. By training the AI on reward-based reinforcement, it adapts to maximize its score while avoiding obstacles and self-collisions.

Using PyTorch, OpenAI Gym, and NumPy, the model learns through experience replay and a Q-learning approach, optimizing performance over time. The training process involves reward shaping, penalizing negative moves while reinforcing positive actions, leading to gradual improvement in gameplay strategy.

This project explores the application of reinforcement learning in game environments, demonstrating the power of AI in learning complex strategies without predefined rules.

Link: github.com/zorogotty14/Snake-game-deep-learning

I developed a system to detect and explain political bias in news content using Large Language Models (LLMs). Leveraging the BIGNEWSALGN dataset, our tool classifies biases (left, lean left, center, lean right, right) and provides detailed explanations, combating echo chambers in today's information landscape. Users can input queries, articles, or URLs, processed through techniques like TF-IDF vectorization and web scraping. Despite computational challenges, our fine-tuned Gemma 7b model achieved 54.50% accuracy for 5-label classification and 67.50% for 3-label classification (left, center, right).

Link: github.com/zorogotty14/IR-CS5604

Selfie-Less Acts is a social media application designed to showcase and share pictures of selfless acts. Users can upload these images to the server and view them via an integrated mobile app. The server is accessed using RESTful API calls.

The project leverages Amazon Web Services (AWS) features such as EC2 instances, security groups, target groups, and an Elastic Load Balancer to build a robust infrastructure. Additionally, we developed a custom orchestrator engine to efficiently handle user requests. Key features of this project include:

• Container Management: Containers can be manually started.
• Load Balancing: Requests are distributed across containers using a round-robin algorithm for improved performance and reliability.
• Fault Tolerance: Continuous health checks monitor the containers to ensure optimal functionality.
• Scalability: The number of act containers automatically scales up or down based on traffic volume managed by the orchestrator.

This architecture ensures efficient, reliable, and scalable service delivery for the Selfie-Less Acts application.

Link: github.com/zorogotty14/SelfieLessActs

Our project, "Gastro-intestinal Disease Prediction," aims to develop a non-invasive, image-based machine learning tool to classify GI diseases. Utilizing models like VGG16, VGG19, ResNet50, DenseNet121, MobileNet, and EfficientNet, we analyze endoscopic images from the KVASIR dataset. Our models reduce the need for invasive procedures, enhancing diagnosis accuracy and accessibility. Each model demonstrates unique strengths, with misclassifications primarily occurring between visually similar classes. Our approach promises to improve healthcare efficiency by supporting faster, more accurate GI disease diagnoses, ultimately benefiting patient care. Future work will focus on optimizing model performance and exploring ensemble methods.

Link: github.com/zorogotty14/Gastro-intestinal-Disease-Prediction

In the AnimePlay project, I demonstrated the separation of the model, view, and controller (MVC) using three Django apps: AnimePlayApp, AnimePlayAuthApp, and AnimePlayActionsApp. This project features user authentication, dynamic content loading, and CRUD operations for anime data. Users can add, edit, delete, and comment on anime entries. Admins have enhanced permissions, including role management and access to all user data. The project integrates Django messaging frameworks, AJAX for dynamic content, and external APIs for enhanced recommendations. Activity feeds and comment threading are also implemented to enrich user interaction.

Link: github.com/zorogotty14/AnimeWebsite

Developed a full-stack web application for a bookstore using Vue.js for the frontend and Java for the backend, implementing RESTful APIs. Incorporated features such as placing orders, generating order details, and managing cart functionality. Utilized DAO classes and model classes for database interaction. Ensured data persistence with local storage for shopping carts. Employed Pinia stores for state management. Styled the interface with CSS Grid, adhering to design and usability principles. Achieved functionality including order placement, cart management, and seamless navigation between pages. This project showcases proficiency in frontend and backend development, database integration, and user experience design.

Link: github.com/zorogotty14/BookStoreWebsite

Developed a sophisticated Android application, FancyGallery, as a culmination of mobile app development coursework. Utilized advanced features including a bottom navigation bar for seamless navigation, REST API integration via HTTPS for fetching curated images from Flickr. Implemented JSON deserialization into Kotlin objects for efficient data handling. Employed RecyclerView to display images in a scrolling grid and integrated cache refreshing functionality. Enabled viewing Flickr web pages for selected images via WebView with progress bar indication. Incorporated MapView to display images on a world map, with customizable markers and handling user interactions. Demonstrated proficiency in mobile app development, networking, UI/UX design, and data management.

Link: github.com/zorogotty14/CS5254-MobileDevelopment