AWS AIP-C01 Certification Exam Sample Questions

AIP-C01 Braindumps, AIP-C01 Exam Dumps, AIP-C01 Examcollection, AIP-C01 Questions PDF, AIP-C01 Sample Questions, Generative AI Developer Professional Dumps, Generative AI Developer Professional Official Cert Guide PDF, Generative AI Developer Professional VCE, AWS Generative AI Developer Professional PDFWe have prepared AWS Certified Generative AI Developer - Professional (AIP-C01) certification sample questions to make you aware of actual exam properties. This sample question set provides you with information about the Generative AI Developer Professional exam pattern, question formate, a difficulty level of questions and time required to answer each question. To get familiar with AWS Certified Generative AI Developer - Professional exam, we suggest you try our Sample AWS AIP-C01 Certification Practice Exam in simulated AWS certification exam environment.

To test your knowledge and understanding of concepts with real-time scenario based AWS AIP-C01 questions, we strongly recommend you to prepare and practice with Premium AWS Generative AI Developer Professional Certification Practice Exam. The premium AWS Generative AI Developer Professional certification practice exam helps you identify topics in which you are well prepared and topics in which you may need further training to achieving great score in actual AWS Certified Generative AI Developer - Professional exam.

AWS AIP-C01 Sample Questions:

01. An insurance company's responsible AI policy states that every customer-facing GenAI release must attach the approved Amazon Bedrock guardrail, must not set temperature above 0.3, and must include the standard AI disclosure in its system prompt. Each release is a configuration file in a repository that names the guardrail ID, the inference parameters and a prompt version from Amazon Bedrock Prompt Management, and an AWS CodePipeline pipeline deploys it. Last month a release shipped without the disclosure because a reviewer missed it.
Which solution will enforce the policy on every release with the LEAST ongoing manual effort?
a) Add a manual approval action in the pipeline and give reviewers a written checklist of the guardrail, temperature and disclosure rules
b) Attach the approved guardrail with a denied topic covering missing disclosures to every model call
c) Turn on AWS CloudTrail for Prompt Management and review new prompt versions every week after release
d) Add a pipeline stage that invokes a Lambda function to check the release configuration and prompt text against each policy rule and fail the stage on any violation
 
02. A bank's internal research assistant runs as a containerized service on Amazon ECS behind an Application Load Balancer and generates long answers with Amazon Bedrock. Employees reach it through a corporate proxy that allows ordinary HTTPS requests but blocks WebSocket upgrade requests.
Answers must appear word by word as the model generates them, and the browser never needs to send data while an answer is streaming.
Which approach meets these requirements?
a) Call ConverseStream in the service and relay each text delta to the browser as server-sent events over the HTTPS response
b) Create an API Gateway WebSocket API that receives each question and pushes every ConverseStream text delta back to the browser on the open connection
c) Have the browser poll a status endpoint every second for the latest saved portion of the answer
d) Call InvokeModel and return the complete answer with chunked transfer encoding on the load balancer
 
03. An insurance company's claims pipeline sends each uploaded claim document to an Amazon Bedrock model with a versioned extraction prompt at temperature 0. Because of client retries and duplicate uploads, about 30% of invocations repeat a document that the pipeline already processed with the same prompt version. Extracted amounts must match the source document exactly. The prompt is revised every few weeks, and after each revision every document must be processed with the new version.
Which caching design will reduce model invocations while meeting these requirements?
a) A semantic cache that returns the stored extraction when a new document's embedding is close to an earlier one
b) Cache results keyed on a hash of the document bytes, its file name and the uploader
c) Cache results keyed on a hash of the document bytes, the prompt version and the model ID
d) Amazon Bedrock prompt caching with the versioned extraction instructions placed first as the cached prefix of every request
 
04. An accounts-payable platform extracts 14 fields from scanned supplier invoices with a large FM on Amazon Bedrock. A test on 2,000 labeled invoices showed that a much cheaper model from a different provider extracts every field correctly on about 85% of invoices. The platform's validation code already detects a failed extraction, such as a missing field or line items that do not add up to the total. Invoice complexity cannot be predicted before processing, and the process must stay fully automated.
Which approach will reduce cost the MOST without lowering extraction accuracy?
a) Use Amazon Bedrock intelligent prompt routing to pick the cheaper or the large model for each invoice
b) Process each document on the cheaper model first, and retry just the ones that fail the existing checks on the large model
c) Route invoices by supplier, sending suppliers known for long invoices to the large model
d) Move all invoices to the cheaper model and queue failed extractions for review by the finance team
 
05. A market-research firm generates client reports with an Amazon Bedrock knowledge base over studies stored in Amazon S3. Every source document already has a .metadata.json file with origin and license attributes, and the data source has been synced. The application calls the RetrieveAndGenerate API but displays only the generated text. Compliance now requires every cited passage in a report to show its origin and license.
Which solution meets this requirement with the LEAST custom code?
a) Instruct the model in the prompt template to state the origin and license after each passage it quotes in the report
b) Add a metadata filter on license to each retrieval request, limiting results to passages whose license permits reuse
c) Read the metadata on each retrieved reference in the response citations and render it
d) Look up each cited document's S3 object tags with a GetObjectTagging call and print them under the passage
 
06. An industrial equipment distributor's chat assistant answers order questions. Before each question reaches the FM, the application extracts any part numbers so that it can attach the matching order records as explicit fields. Customers mention part numbers next to quantities and model years, often with typos, and newer part numbers follow a revised naming convention. The built-in Amazon Comprehend entity types label these numbers inconsistently. The team has a list of valid part numbers and thousands of past chat messages.
Which approach will MOST reliably extract the part numbers?
a) Train an Amazon Comprehend custom entity recognizer from the part-number list and past messages
b) Add a Lambda function that matches part numbers against the list with regular expressions and exact string lookups
c) Instruct the FM in the system prompt to identify part numbers in each question
d) Use Comprehend key phrase extraction and send every phrase to the model
 
07. A telecom provider already runs three specialized agents for billing, technical support and plan upgrades. Customers currently choose an agent from a menu and often pick the wrong one.
The provider wants a single chat entry point that classifies each message, routes it to the most suitable existing agent and keeps the conversation context as the topic moves between agents. It wants to use an AWS open-source framework.
Which solution meets these requirements?
a) Merge the three agents into one Strands Agents agent that holds every tool and instruction set, then retire the existing agents
b) Place the three agents behind an AWS Agent Squad orchestrator that classifies each message and routes it to an agent
c) Send every message to an Amazon Bedrock intelligent prompt router that chooses a model for it
d) Create Amazon EventBridge rules that match message keywords and invoke the matching agent
 
08. A developer is moving a text-generation feature to the Amazon Bedrock Converse API. The model in use supports a top_k sampling parameter, and the team's tests produced the best results with top_k set to 40. The developer's first Converse request fails with a ValidationException. Its inference settings are:
"inferenceConfig": {"maxTokens": 800, "temperature": 0.4, "topK": 40}
How should the developer pass the top_k setting?
a) Replace topK in inferenceConfig with topP set to 0.40
b) Keep topK in inferenceConfig and switch to a Converse-compatible model ID for the same model family
c) Put top_k in the additionalModelResponseFieldPaths array of the request, next to the other sampling values
d) Move top_k out of inferenceConfig into additionalModelRequestFields, using the name the model defines
 
09. A property-management company ran a two-week proof of concept on Amazon Bedrock for an assistant that answers tenant maintenance questions. One developer tested 200 real tenant questions one at a time, and answer quality and response time both met the targets. In production the assistant must handle the evening peak, when several hundred tenants write in at once. The steering committee will approve the full build only if the proof of concept shows that the targets will hold in production.
What should the team add to the proof of concept before the decision?
a) A comparison of the same 200 questions against two more models from different providers
b) A larger set of tenant questions, run sequentially by the same developer, to tighten the confidence interval on the measured answer quality
c) A load test at the expected peak concurrency that measures latency and throttling against the account's Bedrock quotas
d) A Provisioned Throughput purchase for guaranteed production latency
 
10. An HR team hosts a small open-weight text model that drafts answers to employee policy questions. The model runs on CPU and fits in a few gigabytes of memory. Traffic is close to zero for most of the day, but every weekday between 08:00 and 10:00 hundreds of employees use the feature at once.
Cold-start delays are not acceptable during the morning peak, although an occasional slow first response at other times is fine. The team wants to minimize cost while idle and does not want to manage instances.
Which hosting option meets these requirements?
a) A SageMaker AI Serverless Inference endpoint with on-demand capacity only, sized at the maximum memory setting
b)A SageMaker AI Serverless Inference endpoint with scheduled Provisioned Concurrency for the morning peak
c) A SageMaker AI asynchronous inference endpoint that scales out from zero when the morning requests arrive
d)  A SageMaker AI real-time endpoint with a fixed count of CPU instances sized for the morning peak

Answers:

Question: 01
Answer: d		 Question: 02
Answer: a		 Question: 03
Answer: c		 Question: 04
Answer: b		 Question: 05
Answer: c
Question: 06
Answer: a		 Question: 07
Answer: b		 Question: 08
Answer: d		 Question: 09
Answer: c		 Question: 10
Answer: b

Note: Please update us by writing an email on feedback@vmexam.com for any error in AWS Certified Generative AI Developer - Professional certification exam sample questions

Rating: 4.8 / 5 (39 votes)