A company is using a foundation model from Amazon Bedrock for an AI search tool and wants to improve its accuracy by fine-tuning it with the company's proprietary data. Which approach will enable the successful fine-tuning of the model?