Srija Anasuri

Architected a real-time fraud detection platform processing high-volume transaction streams with sub-second decision latency, reducing fraud exposure by $90K through scalable risk scoring and distributed system design. Built event-driven fraud microservices in Java (Spring Boot) with idempotent processing for safe retries and duplicate event handling, improving detection accuracy by 45% and reducing false positives by 30%. Implemented gRPC service-to-service communication to reduce risk evaluation latency by 25%. Managed AWS infrastructure using EC2, EKS, IAM and Terraform for automated scaling and high availability. Designed hybrid persistence using PostgreSQL for ACID and DynamoDB for high-throughput fraud signal storage. Built asynchronous pipelines with AWS SQS to decouple services and ensure resilient, failure-safe message delivery. Developed an AI-assisted fraud reasoning system using LangChain for contextual decision support in production workflows.

Architected a real-time fraud detection platform using Java (Spring Boot) to process high-volume transaction streams with sub-second decision latency, reducing fraud exposure by $90K. Built event-driven fraud microservices with idempotent processing to safely handle retries and duplicates, improving detection accuracy by 45% while reducing false positives by 30%. Implemented gRPC service-to-service communication for low-latency internal risk evaluation, reducing scoring response time by 25%. Managed AWS infrastructure (EC2, EKS, IAM) using Terraform for Infrastructure as Code, enabling automated scaling, high availability, and fault-tolerant microservices execution. Designed hybrid persistence using PostgreSQL (ACID transactions) and DynamoDB (high-throughput fraud signal storage). Built asynchronous processing pipelines using AWS SQS for decoupled execution and reliable retries under financial workloads with failure-safe handling. Developed an AI-assisted fraud reasoning system using

Engineered a distributed payments clearing and settlement system using Java microservices with gRPC communication for near real-time settlement across banking channels, emphasizing strong consistency and high availability. Built asynchronous transaction workflows using AWS SQS to support fault-tolerant settlement processing during peak trading loads with zero message loss. Designed a CQRS architecture separating write and reconciliation read models to increase throughput by ~40%. Developed REST and FastAPI services for settlement validation and reconciliation, reducing mismatch errors by 32%. Architected a multi-database financial system using PostgreSQL, MongoDB, Cassandra, DynamoDB, and Redis with caching to achieve sub-second lookup latency. Implemented centralized observability using ELK, CloudWatch, and distributed tracing to monitor latency, failures, and anomalies across services.