Question 1

A company has a centralized AI team managing Bedrock models in Account A. Three development teams in Accounts B, C, and D need access to a specific Claude Sonnet model with provisioned throughput for their applications. The solution must ensure that these accounts can ONLY invoke this specific model and cannot access any other Bedrock models or features in Account A. The solution must also allow the central AI team to monitor usage and apply guardrails centrally without requiring changes in the development accounts.Which solution meets ALL these requirements?

Options :

A : Enable Bedrock model access in Accounts B, C, and D. Use AWS Organizations service control policies (SCPs) to deny access to all Bedrock models except the Claude Sonnet model. Configure each account to invoke the model directly using their own IAM permissions.

B : Create a cross-region inference profile for the Claude Sonnet model in Account A. Share the inference profile ARN with Accounts B, C, and D. Create IAM roles in each development account with permissions to bedrock:InvokeModel on the specific inference profile ARN only.

C : Create IAM roles in Account A that trust Accounts B, C, and D. Grant these roles permissions to bedrock:InvokeModel with a resource condition limiting access to the specific Claude Sonnet model ARN. Configure applications in development accounts to assume these roles.

D : Create a cross-account inference profile for the Claude Sonnet model in Account A. Use AWS RAM (Resource Access Manager) to share the inference profile with Accounts B, C, and D. Create resource-based policies on the inference profile that allow bedrock:InvokeModel action only from the specified account principals.

Answer: D

Question 2

A biotechnology company has fine-tuned an open-source LLaMA model on their proprietary research dataset to generate scientific hypotheses. They need to deploy this custom model to a production endpoint where it can be invoked by their research application. The team requires control over the instance type (GPU-optimized) and auto-scaling policies.Which AWS service should they use to deploy and host their custom fine-tuned model?

Options :

A : AWS Lambda with container images

B : Amazon Bedrock with custom model import

C : Amazon SageMaker AI with real-time endpoints

D : Amazon EC2 with manual model deployment

Answer: C

Question 3

A fintech company is launching a GenAI-powered "Spending Analyzer" feature. The application uses a Large Language Model (LLM) on Amazon Bedrock to explain transaction history to users. The project has two critical governance requirements:Compliance: The model must strictly refuse to answer any questions related to "Tax Advice" or "Legal Liability," as the company is not licensed to provide these services.Agility: The non-technical Product Team needs the ability to update the "Persona" and "Output Format" (e.g., switching from paragraphs to bullet points) of the analyzer weekly without requiring the engineering team to deploy new application code.Which combination of AWS GenAI features should the developer implement to meet these requirements? (Choose two)

Options :

A : Fine-tune the Foundation Model using Amazon SageMaker JumpStart on a dataset of "Refusal" examples. Train the model to recognize tax-related questions and output a standard disclaimer.

B : Implement Amazon Bedrock Prompt Management. Create a parameterized prompt template that defines the System Prompt (Persona) and output structure. Grant the Product Team permissions to iterate, test, and version the prompt within the Bedrock console.

C : Implement a post-processing AWS Lambda function. Use a keyword list (e.g., "tax", "audit", "lawyer") to scan the generated text. If a restricted keyword is found, overwrite the response with a hardcoded error message.

D : Use AWS Systems Manager Parameter Store to hold the System Prompt string. Update the backend Lambda function to fetch the latest prompt version on every invocation. Give the Product Team access to the Systems Manager console to edit the string.

E : Create an Amazon Bedrock Guardrail. Configure a "Denied Topics" policy that explicitly defines "Tax Advice" and "Legal Advice" with description examples. Attach this guardrail to the model invocation to intercept and block restricted queries.

Answer: B,E

Question 4

A legal technology company is building a generative AI application to draft contract clauses. The application must utilize a repository of 200,000 proprietary legal documents that is updated with new case files every night.The Chief Legal Officer has set the following strict requirements:Attribution: Every generated clause must include a direct citation to the specific case file it was derived from.Freshness: The model must incorporate data from documents added to the repository within the last 24 hours.Accuracy: The solution must minimize hallucinations and strictly adhere to the facts present in the proprietary documents.Which architectural strategy should the GenAI engineer implement to meet these requirements?

Options :

A : Use Amazon Bedrock Custom Models to fine-tune a base Large Language Model (LLM) on the proprietary documents. Configure an automated pipeline using Amazon SageMaker Pipelines to re-train and deploy a new model version every night.

B : Perform Continued Pre-training on an Amazon Titan model using the proprietary documents to adapt the model to the legal domain. Use the resulting custom model to generate clauses based on its internal knowledge of the dataset.

C : Use Amazon SageMaker JumpStart to deploy an open-source model to a real-time inference endpoint. Implement a LangChain workflow that loads the entire document dataset into the model's context window for each inference request.

D : Implement a Retrieval Augmented Generation (RAG) architecture using Amazon Bedrock Knowledge Bases. Configure the data source to sync with the document repository daily and set the generation configuration to return source citations.

Answer: D

Question 5

A media streaming company currently runs a self-managed vector database on Amazon EC2 instances for their content recommendation system. The database stores 150 million video embeddings and handles approximately 50,000 similarity searches per day during peak hours, with search activity dropping to near zero during off-peak nights.The infrastructure team spends significant time managing database scaling, performing backups, applying security patches, and monitoring cluster health. Monthly costs include $8,000 for always-on EC2 instances plus operational overhead. The CTO wants to reduce infrastructure management burden and optimize costs for the variable workload pattern.Which migration strategy addresses these requirements MOST cost-effectively?

Options :

A : Deploy Amazon OpenSearch Serverless with vector engine capabilities. Configure auto-scaling policies to scale compute capacity down to minimum during off-peak hours while maintaining search availability.

B : Containerize the existing vector database using Amazon ECS on AWS Fargate. Implement scheduled scaling policies to stop tasks during zero-activity periods and restart them before peak traffic begins.

C : Migrate the vector database to Amazon RDS for PostgreSQL with pgvector extension. Enable Aurora Serverless v2 auto-scaling to reduce costs during low-traffic periods while maintaining managed database benefits.

D : Migrate embeddings to Amazon S3 vector buckets with vector indexes enabled. Configure S3 Lifecycle policies for index management and leverage the pay-per-request pricing model that eliminates costs during zero-activity periods.

Answer: D