Whenever you're ready, walk me through how you'd approach this.
How would you structure a data pipeline for real-time data processing?
Describe a production-ready design using specific technologies, including ingestion, stream processing, durable storage, serving, orchestration, and observability. Address exactly-once or effectively-once delivery, schema evolution, data quality, late and duplicate events, multi-region resilience, replay, historical backfills, and the tradeoffs between streaming and batch paths.