Databricks Generative AI Engineer Associate

An internal HR assistant must answer questions from policy documents that change every week and cite the source material. Which approach is the best starting point?

• AFine-tune a chat model on past HR tickets
• BBuild a RAG application over the policy documents
• CTrain a sentiment classifier for each policy area
• DHardcode the answers into the system prompt

A claims operations app must answer policy questions, retrieve supporting documents, and create a follow-up task in another system only when needed. Which architecture best fits this use case?

• AA prompt-only chatbot with no retrieval
• BA basic RAG chain with no tools
• CA tool-calling agent with retrieval and workflow tools
• DA fine-tuned embeddings model

You need to convert document chunks into vectors before loading them into Mosaic AI Vector Search. Which model capability matters most?

• AText embeddings generation
• BReranking responses after generation
• CImage generation
• DChat completion

A team decomposed its app into intent classification, retrieval, and final answer generation. Which model-selection strategy usually lowers cost without hurting quality?

• AUse the largest available chat model for every step
• BUse smaller or specialized models for simpler steps and a stronger model only where needed
• CFine-tune a separate model for each step before building a baseline
• DAvoid deterministic preprocessing so the LLM handles everything

You need an LLM to return a stable JSON object with fields `issue_type`, `priority`, and `needs_human_review`. Which prompt change is most effective?

• AAsk the model to be creative
• BProvide the exact JSON schema and instruct it to return JSON only
• CAsk for a markdown table instead
• DRaise the temperature to encourage variety

In a RAG app, users complain that the model confidently answers questions not covered by the retrieved context. Which prompt update is best?

• AAsk for longer and more detailed responses
• BRemove citations to simplify the prompt
• CTell the model to answer only from the provided context and say it does not know otherwise
• DIncrease chunk size until every answer fits in one chunk

You need a chatbot over an employee handbook with citations, but it does not need to take actions, request approvals, or plan across multiple tools. What is the simplest fit?

• AA tool-calling agent with calendar and ticketing tools
• BA fine-tuned classifier
• CA batch AI Functions pipeline
• DA RAG chain over the handbook documents

Where should a Databricks team usually store PDFs and other unstructured source files when building a governed RAG pipeline?

• ADBFS root
• BA notebook output folder
• CUnity Catalog volumes
• DCluster local disk

You try to create a standard Delta Sync vector search index from a Delta table and creation fails immediately. Which missing prerequisite is most likely?

• AThe table was not vacuumed recently
• BChange Data Feed is not enabled on the source table
• CThe table is not Z-ordered
• DThe SQL warehouse is not classic

Why use chunk overlap when splitting long documents for RAG?

• ATo make embeddings smaller
• BTo avoid storing metadata
• CTo eliminate the need for retrieval
• DTo preserve context that spans chunk boundaries

A knowledge base contains long contracts with headings, numbered clauses, and appendices. Which chunking strategy is usually the best first choice?

• AUse fixed 5,000-character chunks that ignore document structure
• BEmbed the entire contract as one vector
• CChunk by semantic or structural boundaries such as sections and clauses
• DSplit every sentence into an isolated chunk with no metadata

A footer saying "Confidential - Company Internal" appears on every page and dominates many retrieved chunks. What is the best preprocessing step?

• ADuplicate the footer in the system prompt
• BRemove repetitive boilerplate before embedding
• CConvert all text to lowercase only
• DSwitch to a larger chat model

Which index type automatically stays aligned with an underlying Delta table as the table changes?

• ADirect Vector Access index
• BSQL warehouse cache
• CDelta Sync index
• DMLflow model registry

You created a Delta Sync index that uses self-managed embeddings and no embedding model is configured. In Databricks Runtime 15.3+, what must you provide for ANN similarity search in SQL?

• AOnly `query_vector`
• BOnly `query_text`
• CNeither, because the index infers the vector automatically
• DOnly `num_results`

When decomposing a GenAI application, which step is usually best handled deterministically before calling an LLM?

• AChoosing an empathetic tone for the answer
• BWriting the final natural-language response
• CDeciding whether the answer sounds persuasive
• DRemoving duplicate documents and normalizing metadata

You want to use `ai_parse_document()` on PDFs stored in a Unity Catalog volume. What input shape does the function expect?

• AA STRING file path
• BAn ARRAY<FLOAT> embedding vector
• CA BINARY column containing document bytes
• DA JSON schema definition

In LangChain on Databricks, what is the typical way to turn a Vector Search index into a retriever?

• AWrap the index with `DatabricksVectorSearch` and call `.as_retriever()`
• BLog the index to MLflow Model Registry first
• CConvert the index into a Delta table and query it only with SQL
• DRun `mlflow.evaluate()` directly on the index

What is the main tradeoff of making chunks extremely small?

• AIt lowers storage cost because metadata is removed
• BIt removes the need for an embeddings model
• CIt can improve precision but increase the risk of missing enough context for generation
• DIt guarantees better groundedness