Yashasvi Shukla

Led a high-performing team of 3 Data Scientists to tackle a wide range of complex use cases and delivered data-driven solutions. Managed project timelines, allocation of resources, and prioritization efforts to ensure the timely achievement of project milestones and deliverables. Maintained transparent and regular communication with senior management and key stakeholders regarding project advancements, outcomes, and valuable insights. Orchestrated the collection, comprehensive study, and interpretation of Call recording. Built sentence transformer model to remove IVR recordings. Developed a robust Multi-label & Multiclass classification model using Bi-LSTM to predict the Intent of the customer, achieving 93% accuracy. Implemented cutting-edge techniques including parallelizing training across multiple GPUs and constructing a three-stage cascaded classifier using XGBoost, TF-IDF, word2vec, LSTM, and fine-tuned BERT. Led a project predicting potential employees eligible for offers related to pregnancy, new babies, mental health, family planning, diabetes and smoking cessation, achieving an F1 score of 95% using RoBERTa. Conducted rigorous data analysis in housing and financial risk data, employing K-means clustering and Explainable AI techniques. Analyzed Market Scanned Healthcare data and developed a predictive model using XGBoost, achieving an F1 score of 87%.

Interpreted customer feedback data and identified the area of improvement by analyzing customer sentiment. Developed an NLP algorithm in healthcare data to automatically route newly opened incidents to application-specific queues, resulting in an 85% reduction in manual effort. Constructed a sentence similarity model to provide suggestions based on issue descriptions, leading to an 11% improvement in Turnaround Time (TAT). Played a pivotal role in analyzing, comprehending, and cleansing chatbot data, implementing a Deep Learning model that achieved a 90% accuracy rate in tagging inputs. Developed an algorithm for the classification of legal documents and implemented an Information Extraction system capable of handling millions of images and PDFs using machine learning and rule-based Named Entity Recognition (NER). Developed an automated tool that streamlined various processes and workflows, resulting in a 95% reduction in manual labor.