Ankita Sarkar

Thesis project: Multimodal traffic data imputation and prediction. • Adapted a spatio-temporal imputation model based on the GNN and RNN for traffic prediction. • Implemented PyTorch-based models for complex time series and geospatial data processing. Skills: Time series data · Pytorch geometric · Graph Neural Network

• Built and deployed ML pipelines using Azure ML, MLflow, and XGBoost/ARIMA, reducing energy imbalance costs by €1.86M through scalable forecasting and KPI-driven evaluation. • Developed real-time data pipelines with Apache Spark and third-party APIs; configured Azure services for scalable processing and analytics. • Automated retraining with drift detection and CI/CD via Azure DevOps; created monitoring dashboards to support cross-functional decision-making. Skills: Azure, Python, SQL, CI/CD, Git, Scrum

• Queried large datasets in Hadoop, Redshift, and DB2, and optimized pipelines and standardized shared formats via API integration. • Built and fine-tuned transformer-based NLP models (e.g., BERT), improving classification accuracy by 5% for insurance use cases. • Performed hypothesis testing and EDA, and created Shiny dashboards and visualizations to drive business decisions from statistical insights. Skills: Python · Hadoop Hive · SQL · Amazon Web Services (AWS) · Jira · Git.

• Developed end-to-end project lifecycles in the life insurance domain, from requirement analysis to deployment, ensuring timely delivery. • Maintained large-scale ETL-style data processing apps on IBM Mainframes (COBOL, JCL, DB2, Shell) for financial operations. • Collaborated with dev, QA, and business teams to deliver scalable solutions with strong data integrity and stakeholder alignment. Skills: Linux · IBM Db2 · IBM Mainframe.