AIP-C01 · Question #7
An ecommerce company is developing a generative AI (GenAI) solution that uses Amazon Bedrock with Anthropic Claude to recommend products to customers. Customers report that some recommended products…
The correct answer is C. Create an Amazon Bedrock Knowledge Bases and implement Retrieval Augmented Generation. The core problem is that the model hallucinates products not in the product catalog (a knowledge grounding problem) and returns irrelevant results. Retrieval Augmented Generation (RAG) with Amazon Bedrock Knowledge Bases directly solves this: it grounds every model response in…
Question
An ecommerce company is developing a generative AI (GenAI) solution that uses Amazon Bedrock with Anthropic Claude to recommend products to customers. Customers report that some recommended products are not available for sale or are not relevant. Customers also report long response times for some recommendations. The company confirms that most customer interactions are unique and that the solution recommends products not present in the product catalog. Which solution will meet this requirement?
Options
- AIncrease grounding within Amazon Bedrock Guardrails. Enable automated reasoning checks. Set
- BUse prompt engineering to restrict model responses to relevant products. Use streaming
- CCreate an Amazon Bedrock Knowledge Bases and implement Retrieval Augmented Generation
- DStore product catalog data in Amazon OpenSearch Service. Validate model recommendations
How the community answered
(55 responses)- A15% (8)
- B7% (4)
- C76% (42)
- D2% (1)
Explanation
The core problem is that the model hallucinates products not in the product catalog (a knowledge grounding problem) and returns irrelevant results. Retrieval Augmented Generation (RAG) with Amazon Bedrock Knowledge Bases directly solves this: it grounds every model response in the actual product catalog by retrieving relevant product records before generation. The model can only reference products that exist in the knowledge base, eliminating recommendations of unavailable or non-existent products. Streaming inference (part of the Bedrock Knowledge Bases integration) addresses the long response time complaint. Option A (Guardrails grounding/automated reasoning) helps with factual accuracy but does not connect the model to the product catalog. Option B (prompt engineering) does not solve hallucination of products absent from the catalog. Option D (OpenSearch + custom validation) is a valid alternative architecture but requires more development and is not the simplest managed solution.
Topics
Community Discussion
No community discussion yet for this question.