Software Development Engineer
Amazon | Bellevue, WA
April 2021 - Present
- Built and operated Python workflows in Apache Airflow (AWS MWAA) to orchestrate reconciliation runs on EMR/Spark, storing outputs in DynamoDB/S3 and analyzing with Athena/SQL.
- Migrated and modernized approximately 28 pipelines and added metrics-based validation to reduce regressions and improve on-call response.
- Developed and maintained Java/Spring Boot microservices for a distributed reconciliation platform with JUnit test suites.
- Improved service reliability and security by mitigating Log4j risk across compute fleets, encrypting sensitive identifiers, and migrating DynamoDB to on-demand capacity to remove throttling-driven incidents.
- Owned end-to-end delivery of a post-payment reconciliation feature across interconnected Java/Spring Boot microservices and REST API integrations with external teams.
- Enabled automated re-evaluation after reimbursement and recovered 160K+ units and $2.5M+ in overpayments within days, supporting $20M+ annualized savings.
- Scaled reconciliation coverage across EU marketplaces by extending decision logic for cross-border warehouse adjustments, improving correctness and delivering approximately $991K annualized savings.
- Expanded automated remediation workflows to identify and correct stranded inventory at scale, processing approximately 250K units/day and driving approximately $3.66M expected savings.
- Shipped an ML-driven item substitution model using AWS SageMaker and productionized it through reconciliation service/data pipeline integrations, helping resolve approximately 1.8M items.
- Built scenario-based ML evaluation and threshold experimentation with scalable data processing (Spark/EMR + workflow automation), reducing manual exception handling from approximately 3.07% to 2.61%.