Prepare and test your skills
Prepare and test your skills
Free sample question, answered and explained below. The Generative AI Leader practice bank has 2,191 more like it. Try 10 free: instant score, no account.
An enterprise customer support team is building an interactive assistant to process incoming equipment repair tickets. The system must meet the following operational and technical requirements:
Which foundation model capability and API configuration should the team select?
Text-only foundation models are specialized natural language processors designed solely to accept and generate textual tokens. They lack the architectural layers required to parse, encode, or interpret visual media such as photographs or diagrams.
Batch prediction is an execution mode designed for processing large volumes of non-interactive data asynchronously in bulk jobs, where turnaround time is measured in minutes or hours rather than milliseconds.
This architecture would be suitable for an offline audit scenario where thousands of historical, purely textual customer tickets are analyzed overnight for reporting trends without any live human interaction.
Single-context recommendation models are purpose-built predictive algorithms (such as retail recommendation engines) designed to suggest items based on a single referenced product interaction on a catalog page.
Greedy token decoding (Top-K = 1) is a sampling setting where the foundation model strictly picks the single highest-probability token at each step during generation, eliminating randomness.
Top-K = 1 makes output deterministic, it does not solve user-perceived latency in conversational interfaces the way token streaming does.This configuration would be selected for an e-commerce product detail page where the goal is suggesting a single complementary product based on the item currently viewed by a shopper.
Multimodal foundation models are generative AI models architected to ingest, process, and correlate diverse data types—such as text, code, and images—within a unified context. Rather than relying on separate specialized pipelines for vision and language, a native multimodal model accepts mixed-modality prompts and generates coherent, contextual natural language outputs.
Streaming response generation (such as utilizing the streamGenerateContent API method) transmits generated output tokens back to the client progressively in real time as they are produced, rather than waiting for the entire generation process to complete.

streamGenerateContent) ensures users see text immediately as it generates, eliminating long perceived wait times during live interactions.Combining native multimodal input processing with streaming response APIs satisfies both the functional data requirements (images plus text) and the human-centric performance requirements (low perceived latency), making it the optimal design choice for interactive enterprise support.
Code-generation foundation models are specialized systems fine-tuned primarily for software development tasks such as writing syntax, translating programming languages, and debugging source code.
Periodic automatic tuning is a lifecycle management feature that retrains or updates model weights at scheduled intervals (such as every three months) using updated historical datasets to maintain relevance over time.
This approach would be appropriate for an automated internal developer tool that generates and maintains software integration scripts and needs regular retraining against updated API repositories.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated