Built and scaled a real-time pricing anomaly detection platform processing 2,500+ TPS using Kotlin/Java microservices and Kafka. Achieved >99.9% decision accuracy and reduced false positives through domain-driven validation and ML scoring orchestration, while cutting P99 latency from 1200 ms to 310 ms by re-architecting synchronous flows into an asynchronous, event-driven pipeline to improve efficiency and reliability at peak scale. Worked in a full-time, fully dedicated contract capacity via Randstad.