Innovative Computer Science Solutions

I am a developer dedicated to the intersection of Computer Vision and Deep Learning. My work centers on building intelligent systems that can perceive and interact with the world in real-time. With a deep focus on Python and frameworks like PyTorch and OpenCV, I build end-to-end pipelines—from data mining and preprocessing to developing neural networks.

Projects

Music Sentiment analysisThis project explores the intersection of NLP and Predictive Modeling within the music domain. It focuses on how text-based user sentiments (emotions) can be used to predict musical characteristics and recommend songs.
I implemented a machine learning pipeline that includes comprehensive Exploratory Data Analysis (EDA), Random Forest classification, and model interpretability using SHAP.
https://colab.research.google.com/drive/1nFDWwa7VTJ-1kPkIcoSHTxh2Ptx2mraz?usp=sharing
Wikipedia RaceThis project explores Graph Theory and Pathfinding Algorithms by programmatically navigating the interconnected structure of Wikipedia to find the shortest link-chain between two disparate topics. By treating Wikipedia as a directed graph where articles represent nodes and hyperlinks serve as edges, the system utilizes a Breadth-First Search (BFS) strategy to guarantee the discovery of the absolute minimum degrees of separation. The implementation involves a robust web-scraping pipeline using BeautifulSoup and Requests to parse HTML and isolate relevant wikilinks while managing an exponentially growing search space. To maintain efficiency, I integrated a visited-set tracking system to prevent infinite loops and memory-optimization techniques to handle high-traffic "hub" pages. The final tool demonstrates algorithmic efficiency and semantic connectivity, proving how even the most unrelated concepts can be linked through a structured, automated traversal of digital information.
https://colab.research.google.com/drive/1laqtZR2yljaOq7-WLFlWB7uA314Ac9I7?usp=sharing
Neural Network from ScratchThis project demonstrates a foundational approach to Deep Learning by implementing a Multi-Layer Perceptron (MLP) and a Linear Regression model from the ground up using NumPy and PyTorch. By constructing the architecture without high-level abstractions, I developed custom implementations for core components including Mean Squared Error (MSE) loss functions, ReLU activation layers, and a manual gradient-descent optimization loop. The project features a comparative analysis between linear and non-linear mappings, showcasing how multi-layer architectures and activation functions enable models to capture complex patterns that simple linear models cannot. Through rigorous training and validation cycles, the project highlights my ability to architect neural networks at the logic level and visualize performance metrics to drive model refinement.
https://colab.research.google.com/drive/1jlG7bEXnI2eABwyGq9Ma_sZGNyPE9pLj?usp=sharing
Convolutional Neural NetworksThis project focuses on the implementation and optimization of Convolutional Neural Networks (CNNs) for image classification using the CIFAR-10 dataset. I developed a modular CNN architecture and conducted a systematic series of experiments to maximize model accuracy, focusing on hyperparameter tuning and data augmentation. By fine-tuning batch sizes and learning rates, and implementing structural enhancements like stacked convolutional layers, I achieved a significant performance boost.
https://colab.research.google.com/drive/1xixPT-mEbpzChuPTHERTWeIXa1p23oJf?usp=sharing 