You are tasked with creating a data pipeline for a social media platform that needs to analyze user interactions. The pipeline must extract data from Amazon Kinesis Data Streams, transform it using AWS Lambda, and load it into Amazon Redshift. How would you design this pipeline to handle the variability in data volume and ensure reliable processing?