Developed an attention-based multimodal model combining voice acoustics (UCI) and gait VGRF signals (PhysioNet) for early Parkinson’s detection.

An end-to-end OCR system for handwritten Hindi word recognition using a CRNN-CTC architecture. The project combines convolutional feature extraction with bidirectional LSTM sequence modeling to recognize handwritten Hindi text without character-level segmentation. Built with PyTorch, the pipeline includes dataset preprocessing, character tokenization, dynamic-width batching, greedy decoding, and evaluation workflows for robust sequence prediction on the IIIT Hindi handwriting dataset.

Developed an AI-powered editorial intelligence platform that predicts news article engagement and optimizes front-page story placement using machine learning and NLP. The system combines headline analytics, semantic embeddings, engagement scoring, and ranking algorithms to dynamically prioritize stories based on predicted audience traction, relevance, and visibility.

Conducted a comprehensive review of deep learning approaches for the early diagnosis of neurodegenerative disorders, focusing on multimodal learning, medical imaging, and predictive healthcare systems. The paper explores recent advances in CNNs, transformers, and hybrid architectures for improving early-stage detection accuracy and clinical interpretability.

Surveyed state-of-the-art machine learning and deep learning techniques for Parkinson’s disease detection using speech, gait, handwriting, and biomedical signals. The review compares traditional ML pipelines with modern multimodal deep learning frameworks, highlighting challenges in interpretability, dataset imbalance, and real-world clinical deployment.