diff --git a/sdk/python/feast/infra/contrib/spark_kafka_processor.py b/sdk/python/feast/infra/contrib/spark_kafka_processor.py
index 32d91b2010..ea55d89988 100644
--- a/sdk/python/feast/infra/contrib/spark_kafka_processor.py
+++ b/sdk/python/feast/infra/contrib/spark_kafka_processor.py
@@ -131,7 +131,7 @@ def batch_write(row: DataFrame, batch_id: int):
             # Also add a 'created' column.
             rows = (
                 rows.sort_values(
-                    by=self.join_keys + [self.sfv.timestamp_field], ascending=True
+                    by=[*self.join_keys, self.sfv.timestamp_field], ascending=False
                 )
                 .groupby(self.join_keys)
                 .nth(0)