Dhruv Sabharwal

Developed a Named Entity Recognition (NER) system using BERT model to analyze call recordings transcripts, encompassing a comprehensive range of 45 predefined labels. Leveraged advanced deep learning libraries such as PyTorch for model development, data loading pipelines, demonstrating proficiency in implementing complex neural network architectures. Utilized the CUDA backend for GPU acceleration, optimizing model training and inference speed, and leveraging parallel computing capabilities. Incorporated the Transformers library from Hugging Face for efficient implementation of BERT-based models, streamlining the process of tokenization and embedding generation. Employed the sklearn library for machine learning utilities, such as multilabel confusion matrix and classification report generation, enabling comprehensive model evaluation and performance analysis. Optimized model training using advanced optimization techniques such as Stochastic Gradient Descent (SGD) and AdamW, fine-tuning hyperparameters to achieve optimal model performance.