Sumesh Pandita

HLM (NLP/LLM) Dashboard:

Python | SQL | Airflow

• Developed modular NLP pipelines in Python and Airflow to automate engagement data processing, cutting manual workload by 40% and reducing turnaround time by 30%.
• Integrated LLMs to convert unstructured text into structured insights, enabling real-time analytics for stakeholders.
• Standardised data transformation logic across pipelines, improving pipeline scalability and maintainability by 50%.

KRI (Risk-Based Quality Management):

Python | R | Airflow | AWS EC2 | S3 | RBQM

• Refactored R-based KRI calculation scripts and implemented parallel execution, decreasing processing time by 60% for
• Automated execution with Airflow and deployed workflows on AWS EC2, improving operational uptime to 99.5% with S3-backed
• Enabled proactive risk monitoring for clinical trials, improving compliance reporting cycle by 25%

Omnichannel Engine:

Python | Kedro | PySpark | MLflow | Airflow | Omnichannel Targeting

• Designed an omnichannel marketing engine using Kedro and PySpark, delivering targeted communications that boosted inactive
• Led the development of a centralized master interaction table, consolidating HCP engagement data from diverse sources to deliver a unified view of campaign performance and engagement effectiveness
• Automated regional model refreshes and implemented MLflow for version-controlled deployments, reducing campaign
• Performed model tuning and registered models using MLflow, enabling traceable, version-controlled ML deployments

• Streamlined legacy GPF operations to enhance data entry accuracy, resulting in a notable 30% improvement.
• Effectively managed State Life Insurance operations for a large workforce of approximately 5,500 employees, overseeing end-to-end administration and compliance processes.
• Developed and implemented strategies to boost employee enrolment in State Life Insurance, ultimately leading to a significant increase of 10–12% through improved data tracking, policy accessibility, and targeted outreach coordination.
• Reduced manual data processing time by 50% through optimised GPF operations, allowing for more efficient use of resources and improved productivity.