A healthcare analytics company has implemented a data ingestion pipeline to process patient monitoring data from wearable devices. The pipeline uses Amazon Kinesis Data Firehose to ingest data into Amazon OpenSearch Service. Currently, the Firehose stream has a buffer interval of 60 seconds, and an OpenSearch dashboard displays real-time alerts about patients' health metrics based on the ingested data. The company needs to optimize the pipeline to achieve sub-second latency for the alerts on the dashboard to respond quickly to critical patient health events. What do you recommend as the most optimal solution?