Generative AI Leader Practice Questions 2 — 50 Questions

Question 1

A bank wants to use LLMs to generate responses for customer support chat. All conversations must be logged, and any PII must be masked. The solution must comply with financial regulations. Which combination of Vertex AI services should be used?

Accepted Answer

Use Vertex AI Agent Builder with Data Governance, which can automatically mask PII and log interactions.. Option D is correct because Vertex AI Agent Builder integrates with Data Governance to automatically mask PII and log interactions, meeting both the logging and compliance requirements without custom development. This managed service ensures adherence to financial regulations by providing built-in data loss prevention (DLP) capabilities and audit trails, unlike the other options which require manual or less integrated approaches.

Answer

Deploy a custom model on Cloud Run and write a Cloud Function to mask PII.

Answer

Use Vertex AI Prediction with a custom container that masks PII before inference.

Answer

Use the Gemini API directly with a custom logging solution in Cloud Logging.

Question 2

A company is building a search application that requires grounding answers in their internal knowledge base. They want to use Vertex AI Search and Conversation with a custom datastore. Which configuration is essential to ensure the model only answers based on their documents?

Accepted Answer

Configure the answer generation to use grounding with the enterprise datastore as the source.. Option C is correct because Vertex AI Search and Conversation provides a built-in grounding capability that explicitly ties answer generation to a specified enterprise datastore. By configuring grounding with the custom datastore as the source, the model is constrained to retrieve and synthesize answers exclusively from the indexed documents, preventing reliance on its parametric knowledge or external sources.

Answer

Enable streaming responses to get real-time answers.

Answer

Fine-tune the model on the company's documents.

Answer

Set the model's temperature to 0 to make responses deterministic.

Question 3

What are THREE best practices for responsible generative AI deployment?

Accepted Answer

Monitor model performance and data drift over time. Option A is correct because continuous monitoring of model performance and data drift is essential for maintaining the reliability and safety of generative AI systems. Data drift occurs when the statistical properties of input data change over time, which can degrade model accuracy and introduce unintended biases. Regular monitoring allows teams to detect these shifts early and retrain or adjust the model to sustain responsible behavior.

Answer

Maximize model size for best accuracy

Answer

Avoid fine-tuning the model to preserve original capabilities

Question 4

A retail company wants to use GenAI to generate product descriptions. They have a small team of data scientists. What is the most efficient approach?

Accepted Answer

Use a foundation model API with prompt engineering and few-shot examples. Option C is correct because using a foundation model API with prompt engineering and few-shot examples is the most efficient approach for a small team. It leverages pre-trained models (e.g., GPT-4, Claude) via API calls, requiring no infrastructure or training data, while prompt engineering and few-shot examples allow the model to adapt to the company's product catalog with minimal effort and cost.

Answer

Collect more data for several months before starting

Answer

Train a model from scratch using their product data

Answer

Buy a proprietary model from a startup

Question 5

A company is evaluating the ROI of a generative AI project. Which metric is most appropriate?

Accepted Answer

Reduction in time to complete tasks using the generative AI tool. Option A is correct because the primary business justification for a generative AI project is operational efficiency, measured directly by the reduction in time to complete tasks. Unlike technical metrics such as model error rate, this metric ties the AI's output to tangible productivity gains, which is the core of ROI analysis in a business context. Generative AI tools are designed to augment human workflows, so time savings translate into cost savings and increased throughput, making it the most appropriate metric for evaluating return on investment.

Answer

Reduction in model error rate on a test set

Answer

Increase in user satisfaction scores

Answer

Cost per inference compared to historical average

Question 6

A financial services firm is developing a GenAI application for investment advice. They need to ensure regulatory compliance. Which business strategy should they prioritize?

Accepted Answer

Implement strict human-in-the-loop review for all investment recommendations. In regulated industries like financial services, GenAI applications must prioritize compliance over speed. Option B is correct because a human-in-the-loop (HITL) review ensures that every investment recommendation is auditable and meets regulatory standards (e.g., SEC or FINRA rules), mitigating risks of hallucinated or non-compliant outputs. This strategy directly addresses the need for accountability and transparency in high-stakes decision-making.

Answer

Rapidly deploy an MVP and iterate based on user feedback

Answer

Open-source the model to gain community trust

Answer

Partner with a cloud provider that offers indemnification for model outputs

Question 7

Which THREE are essential components of a responsible AI strategy for GenAI? (Select three.)

Accepted Answer

Human oversight for critical decisions. Human oversight for critical decisions (C) is essential because GenAI models can produce plausible but incorrect or harmful outputs. A responsible AI strategy mandates that a human-in-the-loop reviews high-stakes outputs, such as medical diagnoses or financial approvals, to prevent automated errors from causing real-world harm. This aligns with the principle of human accountability in AI governance frameworks like the NIST AI Risk Management Framework.

Answer

Use of only open-source models

Answer

Maximum model size

Question 8

A startup with $500k in seed funding wants to integrate GenAI into their SaaS product for automated report generation. They have 2 ML engineers and expect 10,000 monthly users initially. They estimate that using a foundation model API (e.g., Gemini) will cost $0.10 per 1K tokens, and each report uses about 5K tokens. Alternatively, they could fine-tune an open-source model on their domain data, estimated at $50k for compute and $20k for engineering time, with inference cost of $0.02 per 1K tokens on a dedicated endpoint. Which approach is more cost-effective over the first 12 months assuming 50,000 reports per month?

Accepted Answer

Fine-tune the open-source model because it has lower per-report cost. Option D is correct because the total cost of fine-tuning over 12 months is $70,000 upfront plus $0.02 per 1K tokens * 5K tokens per report * 50,000 reports per month * 12 months = $600,000, totaling $670,000. The API approach costs $0.10 per 1K tokens * 5K tokens * 50,000 reports * 12 = $3,000,000, making fine-tuning significantly cheaper at scale despite the upfront investment.

Answer

Use the foundation model API because it has lower upfront cost

Answer

Use a combination of both depending on report complexity

Answer

Build a custom model from scratch

Question 9

A company is deploying a Gemini 1.0 Ultra model for a code generation assistant. They have set up Vertex AI Model Evaluation with a custom evaluation dataset to measure pass@1 accuracy. The initial evaluation shows 65% pass@1. They want to improve to 80% without collecting more training data. They have already attempted basic prompt engineering (e.g., 'write correct code') with limited improvement. Which approach is most likely to achieve the desired improvement?

Accepted Answer

Use chain-of-thought prompting with few-shot examples of correct code generation.. Chain-of-thought prompting with few-shot examples is the most effective approach because it guides the model through step-by-step reasoning, which is critical for complex code generation tasks. This technique leverages the model's in-context learning ability to improve accuracy without additional training data, directly addressing the need to boost pass@1 from 65% to 80%.

Answer

Reduce the temperature to 0 and set top_p to 1.

Answer

Increase the number of output tokens and enable beam search with width 4.

Answer

Apply reinforcement learning from human feedback (RLHF) using a reward model trained on the existing evaluation dataset.

Question 10

A global e-commerce company is using Vertex AI to build a generative AI chatbot for customer support. The chatbot is powered by the Gemini 1.5 Pro model and uses a vector search index for retrieval-augmented generation (RAG) over product documentation. The company has deployed the application in four regions (us-central1, europe-west4, asia-east1, and australia-southeast1) using a multi-region deployment with a global endpoint. The application is critical and requires high availability with a target latency of under 500ms for the RAG pipeline. Recently, users in Australia are experiencing inconsistent latency spikes, with response times exceeding 2 seconds during peak hours. The team suspects that the issue is related to the vector search index's replication and serving configuration. The index has 10 million embeddings with a dimension of 768. It is stored in a single regional bucket in us-central1, and the vector search index endpoint is deployed in all four regions with the same deployed index ID. The team is using the default configuration for index updates and serving. Which action should the team take to resolve the latency issue for Australian users?

Accepted Answer

Move the vector search index to a multi-regional Cloud Storage bucket (e.g., 'us') to reduce latency for index updates.. The latency spike for Australian users is caused by the vector search index being stored in a single regional bucket in us-central1. When the index is updated, the new embeddings must be rebuilt and streamed from us-central1 to the australia-southeast1 endpoint, introducing significant cross-region latency. Moving the index to a multi-regional Cloud Storage bucket (e.g., 'us') allows the index to be served from a location closer to all regions, reducing the update propagation time and improving consistency for Australian users.

Answer

Create a new regional bucket in australia-southeast1 and store a copy of the index there, then redeploy the vector search index endpoint to use the local bucket.

Answer

Deploy a separate vector search index endpoint for each region with its own index copy stored in a regional bucket in that region.

Answer

Increase the number of replicas for the vector search index in all regions to improve throughput and reduce latency.

Question 11

A large enterprise is deploying a generative AI-powered code assistant for their developers. The solution uses Vertex AI with a fine-tuned Codey model. The security team requires that all prompts and responses be logged for audit purposes, but the logs must not contain sensitive information such as API keys or passwords. The operations team is concerned about high latency during peak usage. You need to design a solution that meets security requirements without compromising performance. Which approach should you take?

Accepted Answer

Enable Vertex AI model monitoring with Cloud Logging, and configure a log sink with a custom exclusion filter to redact sensitive patterns before storing. Option B is correct because it uses Vertex AI model monitoring with Cloud Logging to capture prompts and responses, then applies a custom exclusion filter with a log sink to redact sensitive patterns (e.g., API keys, passwords) in real time before logs are stored. This meets the security requirement for audit logging without sensitive data while avoiding the latency overhead of post-processing or a custom proxy, thus satisfying the operations team's performance concern.

Answer

Use Cloud Audit Logs to capture all API calls to Vertex AI, but do not log the actual prompts and responses

Answer

Log all prompts and responses to Cloud Storage and use a Cloud DLP job to scan and redact sensitive data periodically

Answer

Implement a custom proxy that logs all requests after stripping sensitive data, then forward to the model

Question 12

A large enterprise runs a production application that uses the Gemini API on Vertex AI for real-time content moderation. They are experiencing occasional 429 (Too Many Requests) errors during peak hours. Their current quota is 1000 requests per minute (RPM) and they are hitting around 950 RPM on average, with spikes up to 1050. They have already implemented exponential backoff and retry logic. They need to reduce the error rate without reducing the quality of moderation. Which additional measure should they take?

Accepted Answer

Implement a local caching layer for common moderation queries.. Option C is correct because implementing a local caching layer for common moderation queries reduces the number of identical requests sent to the Gemini API, directly lowering the effective RPM without compromising moderation quality. Since the enterprise is already using exponential backoff and retry logic, caching addresses the root cause of hitting quota limits by eliminating redundant API calls, which is a standard pattern for rate-limit mitigation in production AI workloads.

Answer

Deploy the model on a dedicated Vertex AI endpoint with autoscaling.

Answer

Switch to a lower-tier model like Gemini 1.0 Pro to reduce quota consumption.

Answer

Request a quota increase from Google Cloud support.

Question 13

A developer is using the Gemini API to build a chatbot. They want the model to always respond in a friendly, professional tone. Which prompt engineering technique should they use?

Accepted Answer

Set system instructions to 'You are a friendly and professional assistant.'. Option A is correct because setting system instructions is the most direct and reliable way to define the model's persona and behavioral constraints. In the Gemini API, system instructions act as a persistent, top-level directive that influences every response, ensuring the chatbot consistently adopts a friendly and professional tone without requiring repeated examples or parameter tuning.

Answer

Include a few-shot example in every user message.

Answer

Set the temperature to 0.2.

Answer

Set max output tokens to 100.

Question 14

Refer to the exhibit. What is the most likely cause of this error?

Accepted Answer

The user does not have the required IAM role. The error shown in the exhibit is an HTTP 403 Forbidden response, which indicates that the server understood the request but refuses to authorize it. In Google Cloud, this is most commonly caused by the user's identity lacking the necessary IAM role or permission to call the specific API or access the resource. Even if the project ID is correct and the network is functional, a missing IAM role (e.g., `aiplatform.user` or `roles/aiplatform.user`) will result in this exact error.

Answer

The model is too large

Answer

The network is down

Answer

The project ID is incorrect

Question 15

An MLOps engineer wants to implement continuous evaluation of a generative model in production. Which Vertex AI component should they use?

Accepted Answer

Vertex AI Model Monitoring. Vertex AI Model Monitoring is the correct component because it provides continuous evaluation of model performance in production, including detecting prediction drift, data drift, and feature attribution drift. For generative models, it can monitor output quality and safety metrics over time, alerting engineers to degradation or shifts in model behavior without requiring manual intervention.

Answer

Vertex AI Feature Store

Answer

Vertex AI Prediction

Answer

Vertex AI Pipelines

Question 16

Refer to the exhibit. A developer runs this command but forgets to specify the model name. What will happen?

Accepted Answer

The command will fail with an error. In the context of the `gcloud ai models upload` command (or similar model deployment commands in Vertex AI), the model name is a required positional argument. If omitted, the CLI will fail with an error because it cannot proceed without a unique identifier to register the model in the model registry. The command does not default to any name or prompt interactively; it strictly validates required parameters before execution.

Answer

The command will prompt for a name

Answer

The model will be uploaded with a default name

Answer

The command will succeed but the model will be unlisted

Question 17

A company is building a conversational AI using the Gemini API on Vertex AI. They want to reduce the chance of generating toxic content while still allowing creative and engaging responses for their gaming community. Which TWO safety settings should they adjust in the safety_settings parameter?

Accepted Answer

Enable the 'harm_category' filter for 'DANGEROUS_CONTENT' with threshold BLOCK_ONLY_HIGH.. Option C is correct because setting the 'DANGEROUS_CONTENT' category to BLOCK_ONLY_HIGH allows the model to generate creative and engaging responses for a gaming community while still blocking the most severe dangerous content. This balances safety with creative freedom, as the gaming context may involve simulated conflict or action that is not genuinely harmful.

Answer

Set the threshold for 'SEXUALLY_EXPLICIT' category to BLOCK_NONE.

Answer

Set the threshold for 'SEXUALLY_EXPLICIT' category to BLOCK_LOW_AND_ABOVE.

Answer

Set the threshold for 'HARASSMENT' category to BLOCK_LOW_AND_ABOVE.

Question 18

Which TWO strategies are effective for reducing latency in a generative AI chat application deployed on Vertex AI? (Select 2)

Accepted Answer

Use streaming responses. Option B is correct because streaming responses reduce perceived latency by sending tokens to the client as they are generated, rather than waiting for the full response. This leverages server-sent events (SSE) or chunked transfer encoding to deliver partial results immediately, improving user experience in chat applications.

Answer

Deploy on TPU instead of GPU

Answer

Increase the max output tokens

Answer

Use larger batch sizes

Question 19

A research lab is fine-tuning a large language model on a small dataset of medical records. They observe that the model overfits, memorizing specific patient details and producing outputs that violate privacy regulations. Which technique should they apply to improve generalization and reduce memorization?

Accepted Answer

Apply differential privacy (DP-SGD) during fine-tuning. Differential privacy (DP-SGD) is the correct technique because it directly addresses memorization of sensitive patient data by adding calibrated noise to the gradient updates during fine-tuning. This bounds the model's ability to encode any single individual's information, improving generalization and ensuring compliance with privacy regulations like HIPAA.

Answer

Increase the batch size to 64

Answer

Increase the number of training epochs

Answer

Use early stopping based on validation loss

Question 20

Which THREE factors should be considered when choosing between fine-tuning and prompt engineering for a generative AI task? (Choose three.)

Accepted Answer

Availability of labeled training data. Option A is correct because fine-tuning requires a labeled dataset specific to the target task to adjust model weights via supervised learning, whereas prompt engineering relies on the model's existing knowledge without additional training data. Without sufficient labeled data, prompt engineering is often the only viable approach, as fine-tuning would risk overfitting or poor generalization.

Answer

Cost of API calls per request

Answer

Size of the base model