HARSHAVARDHAN NANDURI

· Developed real-time and scalable credit card fraud detection microservices using Python (FastAPI) and Java (Spring Boot), reducing fraud response latency by 40%.

· Engineered REST APIs for ML model inference using XGBoost and Isolation Forest, improving anomaly detection accuracy to 92% in production.

· Designed and implemented a high-throughput data ingestion pipeline using Apache Kafka and Spark Structured Streaming, processing over 100K transactions per minute for fraud scoring.

· Integrated fraud detection engine with upstream transaction systems and downstream alerting channels via Kafka topics, reducing false positives by 25%.

· Leveraged Redis and Amazon DynamoDB to manage transaction state and user session context, enabling better temporal pattern detection.

· Implemented asynchronous processing using Celery with RabbitMQ to ensure scalable and fault-tolerant event handling with automatic retries.

· Built secure internal APIs to support fraud investigation dashboards, allowing analysts to trace transaction flows, model outcomes, and rule triggers with metadata audit trails.

· Designed and scheduled ETL jobs using Apache Airflow to enrich historical data and label fraud events across PostgreSQL, S3, and Snowflake datasets.

· Applied OAuth2 and JWT authentication with RBAC controls for service-to-service security and controlled access to analyst tools.

· Deployed microservices to AWS ECS Fargate using GitHub Actions and Docker, achieving zero-downtime deployments and full automation.

· Set up proactive monitoring and alerts using Prometheus, Grafana, and the ELK Stack for anomaly detection, latency spikes, and failure patterns.

· Collaborated with data science and compliance teams to align model outputs with AML and FFIEC regulatory standards, ensuring audit-ready and interpretable systems.

· Designed feature flag-based toggling in fraud detection microservices, enabling safe rollouts, quick rollback, and A/B testing of model versions in production.

· Conducted performance benchmarking and load testing using Locust and JMeter, ensuring system stability under simulated peak loads of 1 million transactions/hour.

· Contributed to the architecture review board discussions for backend fraud modules, ensuring adherence to security standards, scalability principles, and internal coding guidelines.

· Led the development of scalable services to detect fraudulent payment activities in real time and actively monitored high-risk merchant behavior across Phonepe’s payment ecosystem.

· Spearheaded the implementation of AMLOCK, a robust financial crime detection and management tool, enabling daily scrutiny of high-volume financial transactions.

· Improved service observability by enhancing monitoring and alerting systems, reducing critical alerts by 87% and increasing system availability to 99.9984%.

· Built a library to auto-generate API documentation by capturing unit test metadata and converting it into OpenAPI specs, deployed via an internal Swagger documentation server.

· Reduced customer support tickets by 55.5% by identifying repetitive merchant queries and building self-serve workflows, significantly streamlining agent-merchant interactions.

· Led backend engineering initiatives including sprint planning, roadmap definition, RFC documentation, and architectural reviews for risk-related services.

· Mentored junior engineers, conducted code reviews and pull request evaluations to maintain code quality, and promoted backend engineering best practices within the team.

· Collaborated cross-functionally with product, compliance, and data science teams to align fraud prevention systems with evolving regulatory and business needs.

· Drove continuous performance optimization efforts for critical APIs handling fraud checks and merchant verification flows, ensuring low-latency and high-reliability service behavior.

· Contributed to the internal risk platform’s extensibility by designing plug-and-play support for dynamic fraud rules and real-time transaction scoring.

· Migrated reconciliation processes of 40+ bank gateways from a monolith to a microservice architecture, reducing reconciliation time from 6 hours to 2 hours.

· Designed an event-driven scheduling mechanism using Redis queues for efficient and reliable job execution.

· Automated the complete reconciliation workflow from file upload to final report generation, eliminating manual interventions and human errors.

· Led the development of an internal reconciliation dashboard for finance teams, enabling secure uploads and one-click report generation.

· Achieved 100% reconciliation coverage by automating previously manual, edge-case scenarios across diverse banking formats.

· Implemented RBAC to restrict access to sensitive operations within the reconciliation system, enhancing internal security compliance.

· Developed a scalable batch processing system capable of handling over 2,000 bank files per day at 1,000 transactions per second (TPS).

· Integrated AWS SQS to prioritize and distribute batch file processing tasks across concurrent workers, improving throughput and scalability.

· Improved database performance by 45% and batch job execution time by 20% through advanced SQL tuning and parallel processing techniques.

· Maintained a notification system processing over 20 million daily events across SMS, email, and push channels with guaranteed delivery.

· Built audit trails and reconciliation logs with metadata tagging, enabling traceability for regulatory audits and internal reviews.

· Wrote unit and integration tests for reconciliation modules using JUnit and REST Assured, maintaining 90%+ test coverage.

· Deployed microservices using Docker and Jenkins to staging and production environments with rollback safety via GitHub Actions.

· Implemented retry and dead-letter queue (DLQ) mechanisms for failed reconciliation jobs, ensuring message durability and transparency.

· Participated in sprint planning, technical grooming sessions, and collaborated with cross-functional teams to prioritize reconciliation automation roadmap.