3.3 Identify Grounding Techniques and Their Use Cases. GENERATIVE-AI-LEADER Practice Questions (Page 4)

Questions 16–20

1. 16expert · hard

   A financial services firm is building an analyst assistant that answers questions using a mix of internal research reports and live market data. The team has two constraints: (1) responses must cite the specific internal report used, and (2) responses must be generated within 2 seconds. The internal reports are stored in a document repository, and live market data is available via an API. Which approach best satisfies both constraints?

   • AGround the model with Google Search to find both internal reports and market data.
   • BUse RAG to retrieve relevant internal reports and include citations, and call the market data API separately to inject live figures into the prompt.
   • CSet a very high max output token count and rely on the model's parametric knowledge.
   • DFine-tune the model on all internal reports and market data snapshots.
   Reveal Solution Discussion Select an answer first
2. 17application · medium

   A healthcare technology company needs to build a custom clinical-decision-support tool. The tool must retrieve patient-specific guidelines from a proprietary FHIR-based data store and pass them to a generative model. The team needs full control over the retrieval logic, chunking strategy, and prompt construction. Which Google Cloud capability should they use?

   • AAdjusting sampling parameters to make the model more deterministic.
   • BGrounding with Google Search to retrieve clinical guidelines from the web.
   • CThe RAG APIs to build a custom retrieval-augmented generation workflow.
   • DAgent Search with a standard enterprise search configuration.
   Reveal Solution Discussion Select an answer first
4. 18foundation · easy

   What is the primary purpose of Google Cloud's pre-built RAG solution using Agent Search?

   • ATo replace the need for any external data sources
   • BTo provide a manual process for building RAG pipelines
   • CTo provide a ready-to-use solution for grounding enterprise data
   • DTo train custom language models from scratch
   Reveal Solution Discussion Select an answer first
5. 19expert · hard

   A large enterprise has thousands of internal documents across multiple formats (PDF, Word, slides) and needs to deploy an assistant that answers employee questions. The team has a small ML group but strong software engineers. They need a solution that can be deployed in weeks, not months, and must support incremental document updates. Which approach is the best trade-off?

   • AUse Google Search grounding to index the internal documents.
   • BBuild a fully custom RAG pipeline using the RAG APIs from scratch.
   • CFine-tune a foundation model on all documents.
   • DUse Agent Search, the pre-built RAG solution, to index the documents and ground responses.
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   A news aggregation app wants to ground its summarization feature with current events. Which type of grounding data is most suitable?

   • AThird-party data
   • BFirst-party enterprise data
   • CInternal database
   • DWorld data
   Reveal Solution Discussion Select an answer first