Practice Free NCP-AI Exam Online Questions
Here is the text converted from the uploaded image:
An Accounting Department is thrilled with the RAG Application that the Application & Data Science Teams recently rolled out. However, they provided some feedback that sometimes (approximately 20% of the time), the documents retrieved are not relevant to their prompts or are too generic.
During development, there was extensive testing between models to make sure the best possible model was selected. The Accounting Department emphasizes that when the responses use the right documents, the results are very good and they are pleased with the completeness, accuracy, and coherence of those responses.
What would be a way to address the irrelevant RAG results without having to rebuild the entire workflow?
- A . Replace the embedding model with a larger, more general-purpose language model to improve document retrieval.
- B . Fine-tune the Large Language Model on a broader dataset to enable it to generate more relevant responses.
- C . Implement a rerank model as a post retrieval step to re-order initially retrieved documents based on query-document relevance.
- D . Significantly expand the document knowledge base by ingesting a much larger volume of financial
reports.
The administrator observes a "Pod Network Unreachable" health check failure in the NAI, specifically affecting several application pods on a newly added worker node. Existing pods on other nodes are functioning normally.
Upon initial investigation, kubectl get pods -o wide shows the affected pods stuck in ContainerCreating or Pending status, and kubectl describe pod <pod-name> events include messages, such as:
"FailedCreatePodSandBox: failed to create sandbox: rpc error: code = Unknown desc = failed to set up network for sandbox…" or "network is not ready"
What is the reason for this error and the appropriate next step that the administrator should take?
- A . The Docker/containerd runtime on the worker node is corrupted; reinstall the container runtime.
- B . The kube-proxy service is misconfigured; restart all kube-proxy pods on the cluster.
- C . The underlying Nutanix network segmentation is incorrect; verify VLAN or IP pool configurations are correct in Prism Central.
- D . The NKP CNI plugin on the affected worker node is not functioning correctly; inspect the logs of the CNI poD.
What is the minimum number of supported GPUs required to deploy Google pre-validated LLMs?
- A . 1
- B . 2
- C . 3
- D . 4
An AI/ML administrator has received a message from the cloud platform operations team who manage the underlying compute infrastructure that there may be a resource consumption issue impacting the workload.
The AI/ML administrator isn’t aware of any problems reported from the consumers of the Nutanix Enterprise AI system but has noted that additional workloads were placed on the platform recently, as well as the introduction of GPUs.
With the cloud platform team reporting a resource consumption issue, and the consumers of the service not reporting any issues, what steps should the AI/ML administrator take?
- A . Ask the Cloud Platform Operations Team to add additional CPUs to the Nutanix Enterprise AI Servers.
- B . Increase the Endpoint instances for all of the configured Endpoints on the Nutanix Enterprise AI platform.
- C . Review the Cluster & Node Usage to identify if a specific resource is being exhausted and increase it.
- D . Ask the Cloud Platform Operations Team to scale the Kubernetes Cluster worker nodes up/out.
An NAI administrator has successfully imported a model from Hugging Face and created an endpoint for the model. The endpoint is in the Active state. From within the Endpoint section in NAI, the endpoint has been tested with a Sample Request, the response is accurate, and the Status shows Succeeded.
The administrator has provided the endpoint URL and generated and provided API keys to the developers. However, the developers are having issues connecting to the endpoint. They keep getting 400 Bad Request errors when attempting to prompt the model.
What should the administrator do next to ensure the developers are able to successfully prompt the model?
- A . Provide the developers the ‘Sample Request’ code from the endpoint by clicking ‘View Sample Request’.
- B . Check the Status of the endpoint again to make sure it was not accidentally Hibernated.
- C . Re-import the model from Hugging Face, create a new endpoint and provide new URL and API keys to the developers.
- D . Add additional resources to ‘scale up’ the number of instances of the endpoint.
An administrator is working with a development team to integrate a customer service application with a large language model deployed on the Nutanix Enterprise AI platform. The model provides natural language responses, and the secure application needs to send user input and receive model-generated replies in real time.
The following requirements must be met:
The application must be connected to the correct API endpoint exposed by the model.
A valid API key must be used to authenticate each request.
The response format must match the expected schema in the application.
The administrator must test and confirm successful communication before production rollout.
Which step should the administrator perform to configure and validate the application successfully?
- A . Confirm the response meets requirements using the copied endpoint URL and API key from the portal.
- B . Expose the model endpoint publicly by assigning an external IP address and bypass authentication for faster testing.
- C . Configure access to the endpoint by assigning the application a role in Prism Central.
- D . Set up the application in the hypervisor management interface and validate access using VM-level health checks.
An administrator has been requested by a developer team to provide them with an endpoint.
Which action must the administrator ensure is taken before creating the first endpoint?
- A . Import a Large Language Model (LLM) to Nutanix Enterprise AI and ensure that the import status is Uploaded.
- B . Create a new Nutanix Enterprise AI instance, and ensure that status is Running.
- C . Import a Large Language Model (LLM) to Nutanix Enterprise AI and ensure that the import status is Active.
- D . Create a new Nutanix Enterprise AI instance, and ensure that status is Completed.
An administrator is troubleshooting a slow response from an inference endpoint used by multiple separate applications. The administrator has observed a huge increase in requests.
In order to identify which application is generating the huge increase in requests, what could the administrator do?
- A . Deactivate API keys one at a time.
- B . Increase the assigned compute resources.
- C . Reduce the assigned compute resources.
- D . Deactivate the inference endpoint.
An administrator attempts to import a pre-trained model into Nutanix Enterprise AI (NAI), but the model fails to appear in the deployment list, and endpoint creation is unavailable.
Which action should the administrator take to troubleshoot the issue?
- A . Verify that the model format and metadata meet NAI requirements.
- B . Assign a public IP address to the model container for visibility.
- C . Modify the API key to allow administrative access to all endpoints.
- D . Restart the AI cluster to refresh the model registry and endpoint manager.
An administrator has been asked to install Nutanix Enterprise AI and is defining the NFS storage class for storing models.
What type of access mode needs to be defined?
- A . ReadWriteOnce (RWO)
- B . ReadWriteMany (RWX)
- C . ReadOnlyMany (ROX)
- D . ReadWriteOncePod (RWOP)