Generative AI Leader Practice Test 35 — 15 Questions

Question 1

A company is deploying a document summarization solution using Vertex AI. They want to minimize cost while maintaining quality. Which three strategies should they implement? (Choose THREE)

Accepted Answer

Choose a smaller, specialized model (e.g., Gemini 1.5 Flash). Using caching reduces repeated processing, choosing a smaller model lowers token cost, and batching requests minimizes overhead. Prompt compression is not a standard Vertex AI feature, and using the largest model increases cost.

Answer

Use prompt compression to cut token usage

Answer

Use the largest available model for best quality

Question 2

Which THREE of the following are potential risks when deploying generative AI?

Accepted Answer

Hallucinations. Option A (Hallucinations) is correct because generative AI models can produce fluent but factually incorrect or fabricated outputs, which poses a risk when users trust them for decisions or information. Option B (Memorization of sensitive training data) is correct because models may reproduce verbatim or near-verbatim content from their training corpus, potentially leaking PII, proprietary data, or copyrighted material. Option C (Bias and fairness issues) is correct because models learn statistical patterns from training data and can amplify societal biases, leading to discriminatory or unfair outcomes in hiring, lending, or content moderation. Option E (Toxic or harmful content generation) is correct because generative models can be prompted or inadvertently produce hate speech, harassment, self-harm instructions, or other harmful material. Option D (Increased model accuracy) is not a risk but a potential benefit, so it does not belong among the deployment risks.

Answer

Increased model accuracy

Question 3

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.. 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 4

A real-time customer support chatbot using Gemini is experiencing high latency. The team must maintain response quality while improving speed. Which technique should they implement?

Accepted Answer

Use context caching for frequent queries. Context caching reduces latency by storing frequently accessed query responses or intermediate computations, allowing the chatbot to reuse precomputed results instead of reprocessing identical or similar requests through the full model pipeline. This directly addresses high latency in real-time systems while preserving response quality, as the cached outputs are identical to freshly generated ones for the same input.

Answer

Switch to a larger model

Answer

Increase the batch size

Answer

Decrease the temperature

Question 5

Which TWO are advantages of using Retrieval-Augmented Generation (RAG) over fine-tuning?

Accepted Answer

No need to retrain the base model. RAG retrieves relevant external knowledge at inference time without modifying the base model's parameters, eliminating the need for retraining (option A). This contrasts with fine-tuning, which requires updating model weights through additional training cycles. Because the knowledge lives in an external, updatable index rather than in the model weights, RAG is also better suited for rapidly changing knowledge bases (option E): documents can be added, updated, or removed without any retraining, whereas fine-tuning would require a new training run each time the knowledge changes. RAG thus preserves the original model while augmenting its output with up-to-date information.

Answer

Requires less data preparation

Answer

Lower inference latency

Answer

More secure because model weights are not modified

Question 6

A large enterprise runs a generative AI solution serving millions of daily inference requests. To reduce costs, they propose using serverless endpoints (Vertex AI Prediction) with a custom container, but they notice high latency during cold starts. Which strategy best addresses this problem while minimizing cost?

Accepted Answer

Set a minimum number of replicas to maintain a baseline of always-on instances.. Setting a minimum number of replicas ensures that a baseline of always-on instances is maintained, eliminating cold starts for the majority of requests. This directly addresses the latency spike caused by container initialization and model loading in serverless endpoints, while the cost impact is limited to the minimum replicas rather than scaling all instances.

Answer

Upgrade to GPU-accelerated machines for all replicas.

Answer

Implement client-side request batching to reduce the number of inference calls.

Answer

Use prewarmed containers by setting an idle timeout to keep instances alive.

Question 7

Which Google initiative provides a set of interactive, open-source tools to help UX designers and product managers build human-centered AI products?

Accepted Answer

People + AI Guidebook (PAIR). The People + AI Guidebook (PAIR) is Google's initiative that offers a collection of interactive, open-source tools, methods, and case studies specifically designed to help UX designers and product managers create human-centered AI products. It focuses on the human side of AI, providing frameworks for user research, prototyping, and evaluation to ensure AI systems are intuitive, fair, and useful. Unlike technical libraries, PAIR addresses the design and product management challenges unique to AI, making it the correct answer.

Answer

TensorFlow Privacy

Answer

Model Cards

Answer

Datasheets for Datasets

Question 8

A company is moving a GenAI proof-of-concept to production. They need to ensure the system can handle variable traffic and maintain low latency. Which THREE practices should they implement? (Choose 3)

Accepted Answer

Implement response caching for common queries. Response caching stores the outputs of frequently requested queries, allowing the system to serve them instantly without recomputation, which drastically reduces latency for repeated requests and offloads the underlying model. Auto-scaling for the serving infrastructure lets the system dynamically add or remove capacity as traffic varies, preserving low latency during peaks while avoiding over-provisioning during lulls. Monitoring and alerting on latency metrics provides the observability needed to detect degradation early and trigger operational responses before users are affected. Together these three practices address both variable traffic and low-latency requirements in production.

Answer

Reduce the input context length to the absolute minimum

Answer

Use a single, large instance to handle all traffic

Question 9

A company deploying a generative AI assistant wants to allow users to override the AI's suggestions before final actions are taken. Which design pattern does this represent?

Accepted Answer

Human-in-the-loop. The Human-in-the-loop (HITL) pattern ensures that a human reviews and can override the AI's suggestions before any final action is executed. This design explicitly maintains human oversight over critical decisions, preventing fully automated actions that could lead to harmful or unintended outcomes.

Answer

Human-on-the-loop

Answer

Human-out-of-the-loop

Answer

Automated decision-making

Question 10

A company is comparing Google Cloud Vertex AI vs AWS Bedrock vs Azure OpenAI. Their application requires grounding responses with real-time search results from the internet. Which platform's feature uniquely supports this requirement?

Accepted Answer

Vertex AI with Grounding + Google Search. Vertex AI with Grounding + Google Search is the correct answer because it uniquely provides native, real-time grounding against live internet search results via Google Search, enabling the model to retrieve and cite up-to-date information from the web. This feature is directly integrated into Vertex AI's model serving pipeline, allowing responses to be grounded in current, publicly available data without requiring external API calls or custom retrieval logic.

Answer

Gemini API without grounding

Answer

Azure OpenAI with Bing Search grounding

Answer

AWS Bedrock with Knowledge Bases

Question 11

A global e-commerce company uses generative AI to generate product descriptions in multiple languages. They want to ensure consistency across markets while respecting cultural nuances. Which THREE strategies should they adopt?

Accepted Answer

Develop region-specific prompt templates that incorporate local cultural references and legal requirements.. Option B is correct because region-specific prompt templates let the generative AI encode local cultural references, idioms, and market-specific legal requirements (e.g., advertising claims, required disclaimers) directly into generation, which preserves consistency of brand intent while adapting to each locale. Option C is correct because human-in-the-loop review by local marketing teams catches culturally inappropriate phrasing, mistranslations, and compliance issues that an AI model may miss before content is published. Option E is correct because A/B testing per region provides quantitative engagement metrics (click-through rate, conversion rate, dwell time) that let the company iteratively refine prompts and validate that localized content actually resonates. Option A is not appropriate because a single neutral tone strips out the cultural nuances the scenario explicitly wants to respect and can still be perceived as tone-deaf in some markets. Option D is not appropriate because a single global model with a translation layer typically produces literal translations that lose idiomatic and cultural context, and it does not address per-region legal requirements.

Answer

Standardize all descriptions to a neutral tone to avoid cultural issues.

Answer

Use a single global model with a translation layer to convert English descriptions.

Question 12

An organization is using Vertex AI Agent Builder to create a customer service agent. They want the agent to be able to hand off to a human agent when it cannot answer a question. What should they configure in the agent's design?

Accepted Answer

Use a 'Fallback' intent to route to a human. In Vertex AI Agent Builder (Dialogflow CX), when the agent cannot answer or match a user's request, the conversation triggers a fallback/no-match path. To hand off to a human, you configure that fallback path to route to a live agent via fulfillment, webhook, or live-agent handoff. There is no standard built-in 'Escalation' intent in the product; escalation is an outcome you implement, not a specific intent type.

Answer

Configure 'Slot filling' to collect more info

Answer

Implement a 'Confirmation' prompt for the user

Answer

Add an 'Escalation' intent that triggers a human handoff

Question 13

A developer wants to quickly experiment with different foundation models available in Google Cloud. Which tool should they use?

Accepted Answer

Gen AI Studio in Vertex AI. Gen AI Studio in Vertex AI is the correct tool because it provides a unified interface for discovering, testing, and customizing a wide range of foundation models (e.g., PaLM 2, Gemini, Codey, Imagen) directly from Google Cloud. It allows developers to quickly experiment with different models via a web UI or API without provisioning any infrastructure, making it ideal for rapid prototyping and prompt engineering.

Answer

BigQuery ML

Answer

Cloud Console Compute Engine

Answer

Vertex AI Model Registry

Question 14

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

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.. The latency issue for Australian users is likely caused by the vector search index serving from a single regional bucket in us-central1. When users in Australia query the index, the nearest endpoint must read the index data from a bucket far away, leading to high latency during peak hours. Option B solves this by creating a local bucket in australia-southeast1 and storing a copy of the index there. The endpoint in australia-southeast1 can then serve directly from the local bucket, reducing read latency. Moving to a multi-regional bucket (Option A) does not help because the bucket is still in the US, and Australia still has to fetch from a distant region. Option C would also work but is more complex and costlier than simply using a local bucket with the same endpoint. Option D addresses throughput but not the geographic distance causing latency.

Answer

Move the vector search index to a multi-regional Cloud Storage bucket (e.g., 'us') to reduce latency for index updates.

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 15

Refer to the exhibit. A developer executed the command to list endpoints. They notice that two models are deployed to the same endpoint. What is the most likely reason for this configuration?

Accepted Answer

It is a canary deployment with traffic splitting. A is correct because deploying two models to the same endpoint with traffic splitting is a standard canary deployment strategy. In this configuration, a small percentage of inference requests are routed to the new model while the majority go to the stable model, allowing validation of the new model's performance before full rollout. This is commonly supported by Google Cloud's Vertex AI, where you can deploy multiple models to an endpoint and assign traffic percentages to each model variant (e.g., 90% to the stable model and 10% to the canary model).

Answer

The endpoint is misconfigured and will cause conflicts

Answer

The models are from different frameworks

Answer

It is a batch prediction endpoint