Publications
Research & technical work
Combined text generation with audio deepfake technology to create synthesized, human-like quotes. Utilized LSTM, Transformers, and Tacotron2 for comprehensive audio synthesis.
Technologies & Methods:
Developed a PointNet Classifier using PyTorch for 3D object classification and semantic segmentation. Leveraged the ModelNet10 dataset and preprocessed 3D mesh data into point clouds.
Technologies & Methods:
Comprehensive study on automated grading systems featuring auto-assessment, intelligent feedback, question-answering, and sentiment analysis for educational applications.
Technologies & Methods:
Developed predictive ensemble ML models for patient readmission prediction within 90 days due to heart failure. Achieved state-of-the-art results using XGBoost, LightGBM, and CatBoost.
Technologies & Methods:
Comprehensive analysis of BERT and RoBERTa models for sentiment classification and tokenization of social media content, specifically Twitter data.