Biswajit Panda

Led a critical team project to predict customer churn, leveraging diverse structured and unstructured data sources within the BFSI domain.

Executed comprehensive data cleaning and feature engineering for ~50K customer interaction records (audio, text, demographic), enhancing dataset readiness for ML model training.

Performed NLP preprocessing on ~10K chat transcripts and extracted acoustic features from ~100+ call recordings, facilitating advanced textual and audio analysis for model development.

Developed and fine-tuned machine learning models using ensemble methods, achieving a predictive accuracy of ~87% in customer churn prediction.

Collaborated on the successful deployment of the final ML model prototype on Azure cloud, demonstrating its capabilities to key business stakeholders and enabling data-driven decision-making.

Engineered robust ETL pipelines using Pandas and SQL to process multi-source raw datasets, ensuring data readiness for both reporting and machine learning initiatives.

Created dynamic reports and dashboards in Power BI, resulting in a ~40% reduction in manual reporting effort and improved data accessibility for stakeholders.

Designed and implemented feature engineering pipelines that enhanced ML model input consistency by ~25%, optimizing data quality and model performance.

Facilitated REST API integration and backend data delivery, providing AI-ready data to power client-facing solutions and improve data-driven product functionality.