Real-time data pipelines, stream processing, and high-throughput event handling for financial data flows, market feeds, fraud detection, and the trade-offs between Kafka Streams and Apache Flink.
Stream processing for fraud detection and real-time payments: windowing, stateful joins, exactly-once processing, and the trade-offs between Kafka Streams (in-process) and Flink (separate cluster).
SQL on streams for financial services: stateful aggregations, windowing semantics, temporal joins for sanctions screening, schema management, and the operational trade-offs that determine which engine fits your architecture.
End-to-end production design: Feast feature store, sliding-window velocity checks with Flink KeyedProcessFunction, embedded ONNX model scoring, a dispute-driven feedback loop, and the operational patterns that keep sub-100 ms card authorisation SLAs intact under load.
Avro schema design for pacs.008 with decimal logicalType and PDPL classification metadata, schema-registry-maven-plugin CI compatibility checks, schema evolution patterns, and the governance model that keeps 60 Kafka topics from becoming a fragile mesh of implicit contracts.
Flink VelocityCheckFunction with RocksDB state and TTL, exactly-once Kafka sink configuration, Kafka Streams interactive query REST endpoint, tumbling and session window semantics, and the checkpoint sizing guidance that keeps RocksDB from stalling payment detection pipelines.