GPU-accelerated photo enhancement application using state-of-the-art AI models for image super-resolution and face restoration. Optimized for RTX 3070 Ti with CUDA acceleration, featuring Real-ESRGAN for general image upscaling and GFPGAN for specialized face enhancement. Includes automated batch processing and memory management for large images.

Deep learning model for detecting metastatic cancer in histopathologic tissue images using a custom CNN architecture. Achieved 86.89% accuracy with optimized threshold detection. Features ESRGAN-based super-resolution preprocessing and comprehensive hyperparameter tuning.

Machine learning project to predict wine quality based on physicochemical properties using supervised learning techniques. Features multiple model comparisons (XGBoost, Random Forest, Gradient Boosting), advanced data preprocessing, and SMOTE-Tomek for handling class imbalance. Achieved 79.06% accuracy in quality classification.

Machine learning project analyzing wine market segmentation using unsupervised learning techniques. Features comprehensive clustering analysis using K-means and hierarchical methods, with PCA for dimensionality reduction. Delivered actionable insights for retailers and producers based on price, variety, and region clustering.

Deep learning model using LSTM and GloVe embeddings to classify tweets as disaster-related or non-disaster. Achieved 82% accuracy with optimized threshold detection. Features hyperparameter tuning, comprehensive data preprocessing, and detailed performance visualization.

This project explores unsupervised and supervised matrix factorization techniques applied to text classification and movie ratings prediction. It showcases advanced machine learning concepts such as Non-negative Matrix Factorization (NMF) and logistic regression, with thorough comparisons and performance evaluations.

Development of a high-performing AWS DeepRacer model for the AWS DeepRacer Student League, leveraging reinforcement learning techniques, optimized reward functions, and racing line calculations.

Behavior tree-based autonomous controller for TIAGo robot performing pick-and-place operations in Webots simulator. Features computer vision for object recognition, inverse kinematics for arm control, GPS/compass navigation, and LIDAR-based obstacle avoidance. Demonstrates advanced robotics programming with hierarchical behavior trees and sensor fusion.

Java-based educational word puzzle generator that creates engaging word games from user-provided words and clues. Features word tokenization, random puzzle generation, and comprehensive unit testing. Developed as a graduate course capstone project demonstrating SOLID principles and OOP design patterns.

A structured data visualization project exploring the Titanic dataset to uncover patterns and insights. Focused on designing user-centric visualizations using principles from data cognition, perception, and storytelling.

Implementation of foundational quantum computing algorithms using IBM's Qiskit framework. Features quantum teleportation, GHZ state generation, Bell's inequality testing, and Grover's search algorithm. Includes comprehensive testing suite and quantum circuit visualizations.

An interactive, responsive landing page demonstrating advanced DOM manipulation, performance optimization, and modern JavaScript techniques. Built as a final project for the Udacity Front-End Web Developer Nanodegree program, showcasing expertise in responsive design and event-driven interactivity.

A dynamic and responsive blog website themed around cyberpunk culture, featuring advanced HTML, CSS, and responsive design techniques. Developed as a project for the Udacity Frontend Web Development Nanodegree, it showcases skills in modern web design, accessibility, and modular component-based architecture.