Sai Kavyusha P.

Data Engineer @Charles Schwab

US

Designed and deployed a real-time fraud detection system using XGBoost and SMOTE on AWS SageMaker to reduce financial loss.• Built a FastAPI-based microservice to serve ML predictions in production, containerized with Docker, and automated deployment via CI/CD pipeline using Jenkins, enabling seamless model delivery with 60% lower latency.• Developed an end-to-end NLP pipeline using HuggingFace Transformers (BERT) for financialsentiment analysis.• Led the migration of a model retraining workflow to Airflow + MLflow, integrating experiment tracking, version control, and automated validation, which cut deployment time by 70% and ensured continuous model reliability.• Utilized Apache Spark and PySpark to engineer features from over 100 million transaction records, drastically improving training data pipeline efficiency by 45% for credit scoring models.• Architected ETL pipelines using Apache Spark and Airflow to process and transform multi-source financial data, improving model readiness and reducing data latency by 50%.• Implemented a feature store with Redis and Delta Lake for credit risk models, enabling consistent feature reuse and reducing engineering time by 40%.• Productionized LSTM-based anomaly detection models for transaction monitoring, integrating them with Kafka and Spark Structured Streaming for real-time scoring.