• Contributed to the predictive analytics project team focused on reducing 30-day patient readmissions by identifying high-risk patients through advanced data analysis and machine learning models. • Utilized Python (pandas, NumPy) and R for comprehensive data preprocessing, including cleaning, feature extraction, and transformation of large healthcare datasets from Electronic Health Records (EHR) and operational databases.• Implemented complex SQL queries to extract, aggregate, and cleanse large volumes of historical patient data, ensuring data quality and accuracy for model training and validation. • Developed and optimized machine learning models using scikit-learn to predict patient readmissions. Utilized logistic regression, random forests, and gradient boosting techniques, achieving 85% accuracy in readmission predictions.• Created interactive data visualizations and dashboards in Power BI, enabling easy interpretation of predictive model results and key insights for clinical and operational teams. • Leveraged AWS Lambda functions for serverless processing of incoming patient data, enabling real-time data ingestion, transformation, and feature extraction before feeding the datasets into machine learning models.• Implemented predictive alerts and early interventions for high-risk patients, resulting in a 15% reduction in 30-day readmission rates across the healthcare system. • Optimized resource allocation and reduced unnecessary hospital stays, leading to a 20% increase in patient satisfaction scores by improving the overall patient experience. • Ensured strict adherence to HIPAA regulations during data processing and management, maintaining high standards of data privacy and security. • Managed project deliverables and timelines using Agile methodologies, holding regular sprint reviews, and collaborating closely with cross-functional teams (IT, operations, and clinical staff) to refine models and ensure continuous improvement.

Responsible for designing and developing high-quality prompts that drive GenAI models • Engineered AI solutions on AWS (Claude, Titan, Jurassic) utilizing AWS Lambda for serverless execution, S3 for storage and redshift for data warehousing, for prompt optimization, resulting in a 20% reduction in response time and a 15% improvement in the accuracy of AI generated responses • Implemented iterative fine-tuning and parameter experimentation to enhance the effectiveness and efficiency of AI models, leading to better performance and resource utilization, ultimately benefiting the company's operational efficiency and cost-effectiveness • Optimized AI models and processes for internal employees, enabling them to efficiently retrieve necessary information from company documents, such as insurance claim documents, ensuring a 95% precision rate and fulfilling client objectives of innovative productivity tools and improved internal workflow

• Worked on a project aimed at improving credit risk assessment processes for a major financial services client by analyzing historical loan data and predicting default probabilities. • Achieved an 85% accuracy rate in predicting loan defaults, significantly enhancing the client's ability to assess credit risk more accurately and make informed loan approval decisions. • Utilized SQL to extract, clean, and transform large datasets from relational databases (MySQL, Oracle) containing transactional, credit history, and demographic information, ensuring high-quality data for model development.• Enabled a reduction in loan defaults by 18% by integrating model predictions into the loan approval workflow, ensuring that high-risk applicants were more accurately identified and managed. • Improved loan application evaluation efficiency by 25%, reducing the time taken for credit assessments by automating and streamlining the process with predictive analytics. • Developed and optimized logistic regression and decision tree models using Python) to predict the likelihood of loan default.• Incorporated data validation and quality checks in the ETL pipeline, ensuring that the transformed data met accuracy standards and adhered to business rules, which resulted in a 20% reduction in data errors and improved the accuracy of model predictions.• Used Snowflake Streams to track changes in the loan data, such as new loan applications or default records, allowing real-time updates to the credit risk models and dashboards.• Conducted A/B testing to compare the effectiveness of two different credit scoring models (Logistic Regression vs. Decision Trees) in predicting loan defaults• Leveraged Excel PivotTables to aggregate large datasets, calculating average loan amounts, credit scores, and default rates across different borrower segments, which led to 15% improvement in identifying high-risk segments and uncovering trends and anomalies in credit risk profiles.