Jun 2024 – May 2025 · 11 mos · Kolkata, West Bengal, India

• Part of the Gen AI Practice team.
• Developed Core SDK to create AI bots (RAG and Agentic solutions) for multiple insurance use cases, including claims classification, Q&A, and policy recommendations. Integrated various evaluation frameworks such as RAGAS and Trulens into workflows to ensure model performance and reliability.
• Optimized data pipelines leveraging Company’s SharePoint, Confluence along with Azure OpenAI toolkits, Azure AI Foundry, and LangChain libraries , achieving 95% processing efficiency and reducing search response time by 200ms.
• Streamlined deployment processes using Docker and managed CI/CD pipelines via Azure DevOps.
• Developed custom OCR models with Azure AI Document Intelligence to extract medical codes and critical data from claim documents.
• Fine-tuned Hugging Face Transformers for downstream NLP tasks such as sentiment Analysis and Named Entity Recognition (NER).
• Enhanced various ML pipelines by integrating GPT models, improving performance across multiple tasks.

Jun 2022 – May 2024 · 1 yr 11 mos · Kolkata, West Bengal, India

• In my role as a Machine Learning Engineer, I developed advanced machine learning models to optimize CRM processes, leveraging Snowflake Data Warehouse (DWH) as the primary data source for training. I worked extensively with dimension and fact tables to derive actionable insights for business decision-making.
• Key Projects:
• Customer Churn Prediction for FinTech Applications:
• Developed a Random Forest classifier to predict customer churn using features such as contract details, transaction history, user satisfaction scores, and regional data. Applied SMOTE to handle class imbalance and RFE for feature selection. Implemented Isolation forest, XGBoost classifier models for performance comparison . Used metrics like Precision , Recall , F1-score, AUC-PR to evaluate the model. Finally logged models with hyperparameters into MLflow. The model enabled proactive retention strategies by forecasting churn likelihood and optimizing engagement efforts.
• Customer Segmentation:
• Developed unsupervised machine learning models using clustering techniques such as K-Means and DBSCAN to segment customers based on behavioral and transactional data. Engineered features from customer demographics, purchase frequency, and engagement metrics to improve cluster quality. Evaluated clustering performance using metrics such as silhouette score , enabling the business to tailor personalized marketing strategies and enhance customer targeting.
• Impact & Collaboration:
• These models provided actionable insights into customer sentiment, buying patterns, and behaviors, driving data-driven decision-making. The outcomes contributed to improved customer retention, personalized CRM strategies, and enhanced overall CRM performance.

Oct 2020 – May 2022 · 1 yr 7 mos · Kolkata, West Bengal, India

• Built and managed end-to-end ETL pipelines between CMPT and D&A Salesforce, ensuring seamless data flow.
• Developed and optimized the ETL process for Quote-to-Cash reporting, integrating Salesforce data into Snowflake.
• Created and implemented business logic across multiple platforms, including Salesforce to Snowflake, SAP to Snowflake, Snowflake-to-Snowflake, and Salesforce-to-Salesforce.
• Led the successful migration of data from Siebel to Snowflake, overseeing the full process to ensure data integrity and consistency.