Amazon

Built statistical and ML-powered systems in Amazon Finance Automation to detect invoice fraud, duplicates, and anomalies at scale.

Built duplicate and anomaly-detection capabilities for invoice creation and update workflows handling 1 lakh+ requests every day across Amazon’s retail, non-retail, and corporate finance systems.

Implemented 3 anomaly-detection rules using statistical analysis and machine learning, including Isolation Forest. The rules identified over 100M USD in anomalies within the first month.

Reduced anomaly reporting latency from 24 hours to approximately 5 minutes through Gandalf-Veritas integration, giving finance teams near-real-time visibility into suspicious activity.

Built APIs and operational dashboards for monitoring and alerting, and used AWS CDK to provision infrastructure and CI/CD pipelines across multiple services.

Developed microservices using Java, Kotlin, TypeScript, Python, Elasticsearch, DynamoDB, Coral, Smithy, and AWS.