Built real-time and batch data pipelines to support financial transaction analytics using distributed messaging, orchestration tools, and scalable processing frameworks.
Developed robust ETL workflows using Spark and workflow automation platforms, trained and deployed ML models for fraud detection and behavioral insights, and exposed secure APIs using FastAPI.
Automated infrastructure provisioning through IaC practices and deployed containerized services on a managed orchestration platform for scalable, resilient operations.
Collaborated with analysts and data scientists to define ingestion cadence, reporting KPIs, and real-time metrics visualized through React-based dashboards, Amazon QuickSight, and Kibana.
Built cloud-native data warehouse solutions using Amazon Redshift, applying star and snowflake schema designs, partitioning, and sort key strategies to improve query performance by 50%+.
Engineered real-time and batch ingestion pipelines using Amazon Kinesis, Apache Kafka, and AWS Glue, capturing high-volume payment event streams into Amazon S3 for downstream processing.
Played a key role in back-end modernization by building FastAPI-based APIs and migrating legacy Django logic to async-first services with Pydantic models and SQLAlchemy ORM.
Deployed containerized microservices using Docker and managed deployments using CI/CD (GitHub Actions) with pre-deployment PyTest checks and rollback strategies.
Trained and deployed fraud and churn models using SageMaker and implemented NLP pipelines using spaCy and NLTK to classify and tag user support chats and memos.
Monitored health of data pipelines using CloudWatch, CloudTrail, and Slack-based alerting, reducing MTTR and improving SLA compliance.
Built scalable data models in Redshift using dbt and schema strategies like star/snowflake schema, optimizing query latency by 50%.
Created reusable widgets in TypeScript, enhanced with CSS animations and responsive design using HTML5 flex/grid layouts.
Employed Gradle for front-end tooling and Maven to handle hybrid integrations with external Java-based fraud scoring systems.
Maintained dashboard scalability using React Context, Redux, and lazy-loaded components, reducing initial load time by 40%.