Shivam Swarnkar

Developed an NLP engine for clinical documents to detect contextual, missing data, drug dosage, language, spelling, gender, and taboo errors. Contributed to NLP project for assessing trainee notes and providing feedback. Deployed production model on GCP using Cloud Function, Cloud Run, Pub/Sub, Firestore, and logs explorer. Deployed API with Docker container and cron job. Developed an Auto-completion Framework utilizing RNN models (Utilizing LSTM and GRU), achieving an 86% accuracy rate through hyper parameter tuning and data preprocessing. Developed a Contextual Error Detection Framework through GPU-based fine-tuning of a masked language model (LLM), attaining 91% accuracy, reducing perplexity to 1.45, and seamlessly integrating additional NLP techniques. Enhanced Spellcheck Framework to achieve 90% accuracy and significantly improved the recall rates of both the Missing Data detection and Contradictory Statement modules to 80% and 95%, respectively, through optimization and the utilization of ML models and NLP techniques. As a backend developer, utilized advanced Python packages and AI techniques, led feature enhancements, and resolved backend issues. Conducted data cleaning, preprocessing, and EDA for accurate healthcare data insights. Developed post-production monitoring dashboards with Streamlit and Google Data Studio, improving data visualization and user experience.

Leveraged Mathematical Modeling and Machine Learning techniques (e.g., Linear Regression, Random Forest) to seamlessly scale up R&D processes to pilot scale. Built a Statistical Model to predict sticking effect of tablet for production scale with Python and deployed as a web application using Streamlit. Attained an impressive 92% accuracy rate by crafting a potent Multiple Linear Regression model to forecast Drug-Polymer miscibility, bolstering product stability. Spearheaded process optimization by designing Python-based automation tools, including an employee project status tracker, boosting operational efficiency. Development and deployment of a Statistical model for predicting the process parameters in Lyophilization process with Python.