Engineered features from the SwitchBoard datasets to analyze spoken conversations, calculating linguistic, paralinguistic, and acoustic features. The project classifies conversations into high consideration and high involvement categories based on Deborah Tannen's theory, revealing insights like gender-based differences in loudness and pitch patterns.

Developed an innovative system using Large Language Models to generate engaging and relatable content that sparks reader curiosity. The project explores various prompting techniques on Twitter datasets to optimize content engagement and reader interest through AI-driven content creation.

Created a context-aware system for identifying and disambiguating homographs in text at MAGICS Lab, USF. The project focuses on understanding contextual clues to accurately determine the correct meaning of words with multiple interpretations.

Developed an intelligent tutoring system using LangChain and GPT-4 model, designed to provide personalized educational assistance and interactive learning experiences for students.

Implemented an advanced image classification system using Vision Transformers (ViT) on the CIFAR-10 dataset. The project showcases the power of transformer architectures in computer vision tasks, featuring custom training loops, validation monitoring, and early stopping mechanisms for optimal performance.

Developed an innovative music generation system using Facebook's MusicGen model that can create music in the style of specific artists. The system features prompt engineering techniques to generate music with characteristics of artists like Diljit Dosanjh (with Punjabi beats) and Taylor Swift (pop-country fusion).

A comprehensive framework that balances fairness and accuracy in machine learning models, specifically designed for credit scoring systems. The project implements novel optimization techniques to reduce bias while maintaining model performance.

Implementation of neural networks from scratch in Go language, featuring feedforward and backpropagation algorithms. The project demonstrates how to build efficient deep learning systems using Go's concurrent capabilities.

A sentiment analysis system built using DistilBERT for analyzing movie reviews. The project demonstrates fine-tuning transformer models for sentiment classification tasks.

An intelligent question-answering system built using GPT-3.5 and LangChain. The system can process PDF documents and provide accurate answers based on the content.