Skip to main content

NCP-AAI Real Exam Questions

NVIDIA Agentic AI

144 questions available · Page 1 of 15

Updated Exam DumpsVerified AnswersPass Guarantee

Get Complete Exam Dumps
Question 1 Single choice

You're utilizing an LLM to translate complex technical documentation into multiple languages. The translations often lack nuance and fail to capture the original intent.

What's the most effective strategy for improving the quality of the translations?

  1. A

    Providing the LLM with a glossary of key terms, concepts in all languages and the dataset of previously

    translated text.

  2. B

    Training the LLM on a dataset of translated texts.

  3. C

    Providing the LLM with guidance to "translate the documents" without additional guidance, so it can

    use trained knowledge.

  4. D

    Providing the LLM with guidance to translate "with high accuracy" without additional guidance, so it can

    use trained knowledge.

Show answer and explanation

Correct answer: A

Explanation

A glossary of key terms, concepts across languages, and previously translated text gives the LLM domain-specific guidance. That helps preserve nuance and intent better than a generic instruction such as translating with high accuracy.

Question 2 Multiple choice

An AI Engineer is analyzing a production agentic AI system's compliance with responsible AI standards.

Which evaluation approaches effectively identify potential safety vulnerabilities and ethical risks in multi- agent workflows? (Choose two.)

  1. A

    Emphasize latency metrics and throughput performance as key evaluation factors for safety

    vulnerabilities, providing a baseline for operational measures and resource allocation.

  2. B

    Implement comprehensive audit trails using NVIDIA NeMo Guardrails with semantic similarity checks,

    tracking agent decisions across conversation flows and evaluating policy violations through automated compliance scoring.

  3. C

    Use user feedback as a primary signal for risk identification, emphasizing post-deployment

    observations and qualitative experience reports alongside operational monitoring.

  4. D

    Deploy multi-layered evaluation combining bias detection metrics (demographic parity, equalized odds)

    with adversarial testing to probe agent responses for harmful outputs across diverse user populations

Show answer and explanation

Correct answers: B, D

Explanation

Comprehensive audit trails with NVIDIA NeMo Guardrails give traceability across agent decisions and policy checks. A multi-layered evaluation using bias detection and adversarial testing directly probes fairness and harmful-output risks across user groups.

Question 3 Single choice

Your agent's primary task is to collect data from a device management platform API. This API frequently returns data in JSON format, but the underlying schemas can evolve without detailed documentation.

What's the most effective strategy for the agent to maintain consistent access to the data, despite potential schema changes?

  1. A

    Building a JSON schema validator that constantly monitors and enforces the current API schema.

  2. B

    Manually inspecting the API response's structure and updating the agent's code when a change is

    detected.

  3. C

    Building a fixed data extraction method based on the API's historical response patterns.

  4. D

    Building a flexible data mapping layer that can adapt to changes in the API response structure.

Show answer and explanation

Correct answer: D

Explanation

When JSON schemas can evolve without clear documentation, a flexible data mapping layer can adapt field handling as response structures change. Fixed historical extraction or manual code updates are brittle and can break access whenever the API shifts.

Question 4 Single choice

What benefits does a Kubernetes deployment offer over Slurm?

  1. A

    Kubernetes provides autoscaling, auto-restarts, dynamic task scheduling, error isolation with

    containers, and integrated monitoring.

  2. B

    Kubernetes is the best option for both training and inference, offering advantages for resource

    management and workload visibility over traditional HPC schedulers like Slurm.

  3. C

    Kubernetes is more optimized for batch jobs to achieve high throughput, and also provides for

    monitoring and failover in large-scale workloads.

Show answer and explanation

Correct answer: A

Explanation

Kubernetes adds operational features that Slurm is not primarily designed around: autoscaling, auto-restarts, dynamic task scheduling, container-based isolation, and integrated monitoring. Those capabilities help run changing service workloads reliably.

Question 5 Single choice

In a ReAct (Reasoning-Acting) agent architecture, what is the correct sequence of operations when the agent encounters a complex multi-step problem requiring external tool usage?

66

66

  1. A

    Thought --> Answer --> Action --> Observation

  2. B

    Action --> Thought --> Observation --> Action --> Thought --> Observation --> Answer

  3. C

    Observation --> Thought --> Action --> Observation --> Thought --> Action --> Answer

  4. D

    Thought --> Action --> Observation --> Thought --> Action --> Observation --> Answer

Show answer and explanation

Correct answer: D

Explanation

ReAct starts with a thought to decide the next step, takes an action such as using a tool, then reads the observation before reasoning again. Repeating thought, action, and observation until enough information is gathered leads naturally to the final answer.

Question 6 Single choice

A customer service agent sometimes fails to complete multi-step workflows when APIs respond slowly or inconsistently.

Which approach most effectively increases robustness when working with unreliable APIs?

  1. A

    Restrict available tools to reduce decision complexity

  2. B

    Add retries with exponential backoff and set request timeouts

  3. C

    Cache recent API results to limit unnecessary repeated calls

  4. D

    Adjust generation parameters to produce more predictable responses

Show answer and explanation

Correct answer: B

Explanation

Slow or inconsistent APIs need bounded waiting and controlled retry behavior. Request timeouts keep each call from blocking the workflow indefinitely, while retries with exponential backoff give temporary failures time to clear without overloading the API.

Question 7 Single choice

Integrate NeMo Guardrails, configure NIM microservices for optimized inference, use TensorRT-LLM for deployment, and profile the system using Triton Inference Server with multi-modal support.

Which of the following strategies aligns with best practices for operationalizing and scaling such Agentic systems?

  1. A

    Use Docker containers orchestrated by Kubernetes, implement MLOps pipelines for CI/CD, monitor

    agent health with Prometheus/Grafana.

  2. B

    Deploy agents on bare-metal servers to maximize performance and avoid container overhead, using

    manual scripts for orchestration and monitoring.

  3. C

    Deploy all agents on a single high-performance GPU node to reduce latency, and use cron jobs for

    periodic health checks and updates.

  4. D

    Run agents as independent serverless functions to minimize infrastructure management, relying

    primarily on cloud provider auto-scaling and logging tools.

Show answer and explanation

Correct answer: A

Explanation

Docker containers orchestrated by Kubernetes provide repeatable deployment and scalable scheduling for production agents. MLOps pipelines support CI/CD, while Prometheus and Grafana give operational visibility into agent health instead of relying on manual scripts.

Question 8 Single choice

Set which architecture for a RAG agent that needs fresh policy updates during the day while keeping retrieval latency low?

  1. A

    One large service that handles ingestion, retrieval, prompting, and response formatting together.

  2. B

    A synchronous polling loop where the model reads the entire policy store for every prompt.

  3. C

    Loosely coupled services for indexing, retrieval, and prompting with event-driven updates.

  4. D

    A single manually refreshed database with no separate retrieval component.

Show answer and explanation

Correct answer: C

Explanation

Fresh policy changes and low latency are easier to balance when indexing, retrieval, and prompting are loosely coupled. Event-driven updates can refresh the index as policies change, while the retrieval path stays focused on serving relevant chunks instead of scanning the full store each prompt.

Question 9 Single choice

You're building a RAG system that uses RAG Fusion.

Which of the following approaches would be most effective in determining how to combine information from multiple retrieved chunks?

  1. A

    Filtering out chunks considered inconsistent with others before presenting information to the LLM.

  2. B

    Using the LLM to automatically identify the most important sentences within each chunk and combine

    them.

  3. C

    Manually selecting the most relevant sentences from each chunk and inserting them into the LLM

    prompt.

  4. D

    Concatenating the text from all retrieved chunks into a single block to form the response.

Show answer and explanation

Correct answer: B

Explanation

Using the LLM to automatically identify important sentences within each retrieved chunk supports the fusion step: the model can decide what information matters and combine it into a coherent response. Simple concatenation or manual sentence picking misses that adaptive synthesis.

Question 10 Single choice

Failures in a warehouse operations scheduler occur when shipment priorities change after execution starts; which design change best addresses this issue?

  1. A

    Shorten the route sequence so fewer steps can become stale.

  2. B

    Use a goal-driven agent that can revise decisions from context and environment feedback.

  3. C

    Increase parallel execution of the same fixed workflow stages.

  4. D

    Cache the first valid plan and reuse it until the next daily planning cycle.

Show answer and explanation

Correct answer: B

Explanation

When shipment priorities change during execution, a fixed plan becomes stale. A goal-driven agent can revise decisions from current context and environment feedback, so scheduling adapts instead of reusing an outdated route or workflow.