A retailer collects purchase records from 20,000 stores and sends them to Amazon S3 using Amazon Kinesis Data Firehose. Each store runs a small server application that transmits the data over the internet. The data is used to retrain a machine learning model daily. The data science team identified existing fields that should be combined into new attributes before training. Which change will produce the transformed records while adding the least operational overhead?

The correct answer leverages Kinesis Data Firehose's built-in data transformation capability, which integrates directly with AWS Lambda. This approach is serverless, scales automatically with the incoming data stream, and requires minimal operational overhead because Firehose manages the invocation and error handling for the Lambda function. The transformation happens in real-time as data flows through Firehose, ensuring the data is ready for model retraining quickly. Incorrect options: An Amazon EMR cluster is overkill for simple record transformations and introduces significant operational overhead for cluster management and scheduling. Deploying Amazon S3 File Gateway in 20,000 stores is a massive operational burden and doesn't directly address the data transformation requirement efficiently. Provisioning and managing a fleet of Amazon EC2 instances for a daily cron job is less efficient and more operationally complex than a serverless Lambda function integrated with Firehose.