Event sourcing, pub/sub, broker design, schema evolution, and async processing patterns for real-time financial systems — from real-time payments to trading and core banking modernisation.
Where Kafka fits, where IBM MQ still wins, and how to run them side by side. Topology, schema evolution, idempotent consumers, exactly-once semantics, HA & DR with MirrorMaker 2.
Running Strimzi (and Red Hat AMQ Streams) Kafka on private, air-gapped OCP in a regulated bank: operator topology, listeners, OAuth via internal IdP, mTLS with the internal CA, NetworkPolicies, multi-AZ HA, and MirrorMaker 2 DR.
When event sourcing is worth the cost: event store design, CQRS projection patterns, ISO 20022 schema alignment, snapshotting, idempotency guarantees, SAMA audit trail requirements, and the Kafka-backed read model that scales.
Choreography vs orchestration sagas, compensating transactions, idempotent saga steps, and recovery patterns for multi-step IPS payment workflows — including the pivot-point, the dead-letter queue, and the SAMA audit trail.
Kafka @RetryableTopic, IBM MQ AMQMDLQ handler rules, poison message classification, idempotent replay service, ISO 20022 failure routing, and the DLQ observability patterns that surface errors before they become SAMA incidents.