Real-Time Banking: Iceberg, Interoperability, and the Rapid Rise of Flink and Streaming in Financial Services
Lessons from the Banks and Payment Processors Leading the Shift
Banking is going real-time. Fraud stopped mid-transaction, intraday risk, real-time P&L calculations, agents requiring real-time context, and more. Some of the largest banks and payment processors have built around event-driven architectures with Apache Flink at the core. This event is about what they learned and why they chose this architecture.
One lesson surprises people: it was never just about speed.
Banks are expanding Flink even for workloads where minutes of latency would be fine, because Flink is built for stateful, continuous computation. Flink applications remember: every customer's behavior profile, every account's exposure, all updated live with each event. The alternative, re-scanning ever-growing tables in a warehouse or batch platform on a schedule, costs more at every scale point and can't support applications that need to react to each event in context. Query platforms answer questions about your data. Flink builds applications that run on it.
The second trend is Apache Iceberg and interoperability. Financial services companies are converging on open table formats as the shared foundation under every engine: one copy of the data, on cheap object storage, that Flink streams into, and your warehouse, batch, and ML platforms read directly. Lower cost, real interoperability, and no lock-in.
Put together, a common pattern is emerging in financial services: Kafka and Flink for movement and computation, Iceberg as the shared foundation, and your existing engines, whether Spark, Trino, or your warehouse, reading the same copy of the data. Not rip and replace. A better division of labor.
One of the sessions will be presented by Pedro Mázala, Streaming Solutions Engineer at Ververica, who works with engineering teams to design and operate production grade streaming systems built on Apache Flink. Drawing from real customer implementations, Pedro will share how financial platforms build metered billing systems where every number must be accurate, reproducible, and auditable. He'll cover how Flink enables reliable reprocessing at scale without double charging customers, why state management and delivery guarantees matter, and what teams learn when operating these systems in production. The session will also explore the practical realities of running Flink with Apache Iceberg, from checkpointing and compaction to balancing data freshness, performance, and cost as systems grow.
Join the original creators of Apache Flink and practitioners from financial services for talks on which applications drove the shift, what runs where, what Iceberg adoption looked like in practice, what it meant for cost and governance, and what's next with GenAI.
Who should attend: data engineers, software architects, data platform leaders, and engineers building applications.
Great talks, honest Q&A, good food and drinks, and a room full of people who actually get it. That's the evening.
Location: Charlotte City Club
