Screening — Jonas Meyer

At Trade Republic, I designed a payment reconciliation service handling two million transactions a day using idempotency keys and the outbox pattern for exactly-once processing and Kafka for settlement events. I owned it from design through on-call

I ensure consistency with idempotency keys, the transactional outbox pattern, and at least once delivery with deduplication. For cross-service workflows, I use sagas plus nightly reconciliation jobs to catch drift

We had latency spikes under peak load. I traced it to connection pool exhaustion from a slow downstream call, added timeouts, a circuit breaker and bulkheads, and an alert on pool saturation. Latency dropped 40%.