AWS Certified Generative AI Developer - Professional AIP-C01

By Amazon · Question Mode

QuestionQ29

Testing, Validation, and Troubleshooting

A healthcare company uses Amazon Bedrock to deploy an application that produces summaries of clinical documents. The application has inconsistent response quality, including occasional factual hallucinations. Monthly costs are 40% higher than the company projected. A GenAI developer must implement a near-real-time monitoring solution that detects hallucinations, identifies abnormal token consumption, and provides early warnings for cost anomalies. The solution must involve minimal custom development and maintenance overhead.

Which solution meets these requirements?

Explanation

Amazon Bedrock Guardrails contextual grounding checks detect and filter hallucinations by assessing whether generated content is factually grounded in a supplied reference source, including for summarization. Amazon Bedrock provides InputTokenCount and OutputTokenCount CloudWatch metrics, and CloudWatch anomaly-detection alarms identify departures from expected metric patterns, enabling early warning of unusual token-driven costs. Bedrock model invocation logging can retain text input and output data in Amazon S3.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!