How have you implemented Kafka in a project?

In a recent e‑commerce platform, we used Kafka as the backbone for real‑time order processing. Producers were built in Java using the Kafka client library; each order event was serialized with Avro and sent to the "orders" topic. We employed a Schema Registry to enforce schema evolution. Consumers ran as microservices in Docker containers, each subscribing to specific topics (e.g., "orders", "payments") and using the Kafka Streams API for stateful processing. We configured topic replication factor 3 and partition count 12 to balance throughput and fault tolerance. For durability, we set a retention period of 7 days and used log compaction for the "order‑status" topic. Monitoring was handled via Confluent Control Center, and we used Kafka Connect to stream data into downstream systems like HDFS and Elasticsearch. This architecture allowed us to process thousands of events per second with sub‑second latency while ensuring data consistency and recoverability.