Have you implemented Kappa architecture in a project?

Yes, in a log analytics platform we adopted Kappa architecture to simplify our data processing pipeline. Instead of maintaining separate batch and stream pipelines, we used Kafka as the single source of truth and processed all data in real time with Spark Structured Streaming. All events were written to a raw topic, and we applied stateless transformations (filter, map) and stateful aggregations (windowed counts) directly in the stream. For reprocessing, we simply replayed the topic from an earlier offset, eliminating the need for a separate batch job. We also used a changelog topic to store intermediate state, enabling fault tolerance. This approach reduced operational complexity, ensured consistent data views, and allowed us to roll back or recompute metrics by replaying the same stream.