Databricks-GenAI-Assoc Practice Test 16 — 15 Questions

Question 1

A data science team wants to expose a fine-tuned LLM as a REST API for an internal web application. They have registered the model in Unity Catalog and want Databricks to manage the serving infrastructure, autoscaling, and request routing. Which Databricks capability should they use?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Databricks Model Serving, which creates a managed endpoint that autoscales and routes requests to the model.. Databricks Model Serving is the managed capability that turns registered models into autoscaling REST endpoints with routing and load balancing. The other options either run unmanaged code on clusters, execute SQL, or precompute batch outputs, none of which deliver interactive, managed model inference for a web application.

Answer

A Delta Live Tables pipeline that materializes model outputs into a table the web app polls.

Answer

A Databricks job cluster running a notebook that starts a Flask server on the driver.

Answer

A Databricks SQL warehouse with a custom HTTP connection to the model.

Question 2

An engineer deployed a RAG agent to a Model Serving endpoint and enabled inference tables for payload logging. After a week, reviewers notice that logged requests contain customer email addresses and order identifiers. Compliance requires that raw prompts and responses not be stored in plain text. Which change should the engineer make?

Accepted Answer

Keep inference tables enabled but restrict SELECT on the inference table to a compliance group and apply row filters or column masks on the sensitive columns.. Because inference tables are Unity Catalog tables, the engineer can apply column masks and row filters to redact or restrict sensitive fields and grant SELECT only to authorized principals. This preserves the observability inference tables provide while preventing plain-text exposure of personal data, which is the precise compliance outcome required.

Answer

Enable payload logging but write inference tables to a workspace-local path outside Unity Catalog.

Answer

Disable inference tables entirely so no request or response data is persisted.

Answer

Configure the endpoint to log only the model's token counts and latency, then reconstruct prompts from the application's own logs.

Question 3

A platform team is preparing to deploy a GenAI chat application built on Mosaic AI Model Serving. They want the deployment to enforce least-privilege access for the application and to keep the serving environment reproducible. Which two practices should they implement? (Choose two.)

Accepted Answer

Pin the model's dependency versions in the logged MLflow environment so the serving environment installs a known, reproducible set of packages.. Least privilege is achieved by granting the application principal only endpoint query rights plus the minimum Unity Catalog reads for the model and its data. Reproducibility is achieved by pinning dependency versions in the logged MLflow environment so the serving container installs a known package set. Workspace admin, dropping Unity Catalog, and always-latest installs each trade away security or reproducibility, so they do not meet the two stated goals.

Answer

Allow the endpoint to install the latest version of every dependency at startup so the environment always has current patches.

Answer

Give the application principal workspace admin so it can create endpoints and modify cluster policies as needed during deployment.

Answer

Disable Unity Catalog for the model and store it in the workspace model registry to simplify permission management.

Question 4

A team wants to review what a Databricks Asset Bundle deployment would change in the production workspace before it actually creates or modifies any resources. Which command should the engineer run first?

Accepted Answer

`databricks bundle plan -t prod`. The `databricks bundle plan` command produces a deployment plan that lists the resources the deployment would create, update, or delete in the target workspace, without applying any change. Running it with `-t prod` previews the production deployment. This lets the team review the impact before executing `databricks bundle deploy`, which is the recommended safe release workflow.

Answer

`databricks bundle deploy -t prod --dry-run`

Answer

`databricks bundle validate -t prod`

Answer

`databricks bundle summary -t prod`

Question 5

An engineer is designing a RAG application where the LLM must answer questions using only the retrieved context and must refuse to answer when the context is insufficient. Which prompt design approach best enforces this behavior?

Accepted Answer

Use a system prompt that instructs the model to answer only from the provided context and to respond with a fixed refusal phrase when the context does not contain the answer.. The most reliable way to constrain a model to retrieved context is a system prompt that explicitly limits answers to that context and defines a refusal response when the information is missing. This gives the model a clear rule and a fallback, reducing hallucinations. Retrieval volume, temperature, and question repetition do not establish grounding boundaries.

Answer

Increase the number of retrieved chunks to ten so the model always has enough context to answer.

Answer

Set the model's temperature to zero and rely on deterministic decoding to prevent unsupported answers.

Answer

Append the user's question twice in the prompt to reinforce the instruction to stay grounded.

Question 6

An engineer is packaging a custom PyFunc model that wraps an open-source LLM and must deploy it to a Databricks Model Serving endpoint. The endpoint must load the model from Unity Catalog and expose it through a REST API. Which two actions are required to make the deployment succeed? (Choose two.)

Accepted Answer

Register the logged model in Unity Catalog and grant the serving identity USE CATALOG, USE SCHEMA, and EXECUTE or SELECT privileges on the model version.. A successful custom model deployment requires a correctly logged artifact with all dependencies declared, and a Unity Catalog registration with privileges granted to the serving identity. Environment build failures and permission errors are the two dominant causes of endpoint creation failure. Flavor conversion, attached clusters, and scale-to-zero settings are unrelated to whether the deployment itself can complete.

Answer

Set the endpoint's scale-to-zero configuration to disabled so the model stays in memory permanently.

Answer

Attach the model to a Databricks cluster that stays running so the endpoint can proxy requests to it.

Answer

Convert the PyFunc wrapper to a scikit-learn estimator so the serving container recognizes the flavor.

Question 7

A Mosaic AI Agent application deployed as a Databricks App intermittently returns stale answers after the team updates the underlying vector index. The app caches a client to the serving endpoint and an index handle at module import time. Which change best resolves the staleness while keeping latency low?

Accepted Answer

Initialize the endpoint client once but resolve the current index version per request, and invalidate the cached handle when the version changes.. Staleness comes from caching an index handle at import time, so the fix is to keep the cheap-to-reuse client but resolve the index version per request and invalidate the handle when it changes. This preserves the latency benefit of a persistent client while ensuring answers reflect the latest index, avoiding both constant rebuilds and redeploy-on-every-update.

Answer

Increase the app's compute size so the cached handle is refreshed more frequently by the runtime.

Answer

Disable caching entirely and rebuild the endpoint client and index handle on every request.

Answer

Redeploy the app after every index update so the module reinitializes with a fresh handle.

Question 8

A generative AI engineer is building a RAG chain on Databricks using LangChain. The chain must call an external LLM provider via a Databricks Model Serving endpoint that proxies the provider, and the engineer wants to avoid hardcoding credentials in notebooks. Which approach should the engineer use to authenticate calls from the chain to the serving endpoint?

Accepted Answer

Configure the serving endpoint with the provider credential stored as a secret in the endpoint's environment, and call the endpoint using the Databricks SDK or LangChain's Databricks LLM class with workspace authentication.. The serving endpoint is the correct place to hold the external provider credential because it centralizes secret storage, rotation, and auditing. Clients then authenticate to the endpoint with workspace identity rather than the provider key. LangChain's Databricks LLM integration and the Databricks SDK both support this pattern, keeping credentials out of notebooks and enabling Unity Catalog governance over the endpoint.

Answer

Pass the provider API key as a plaintext argument in the LangChain ChatOpenAI constructor and rely on notebook access controls to protect it.

Answer

Use a personal access token generated for a service principal and embed it in the chain's prompt template so the LLM can forward it to the provider.

Answer

Store the provider API key in a Databricks secret scope and reference it with dbutils.secrets.get inside the notebook before instantiating the LLM client.

Question 9

A team is deploying a Databricks RAG chatbot that must serve interactive traffic with low latency while also allowing the data science team to test prompt variations safely. The engineer wants the production endpoint to keep serving stable traffic even while a new prompt template is being evaluated. Which approach best meets these requirements?

Accepted Answer

Create a separate Model Serving endpoint for the experimental prompt and route a small percentage of traffic to it for comparison. Separating the experimental prompt onto its own Model Serving endpoint preserves production stability while enabling controlled comparison. The team can direct a small share of traffic to the experimental endpoint, measure quality and latency, and promote the prompt only after it proves better, all without redeploying or interrupting the production endpoint.

Answer

Lower the production endpoint's concurrency limit and run the experiment on the same endpoint during off-peak hours

Answer

Edit the prompt template in the production endpoint's registered model version and restart the endpoint

Answer

Add the new prompt as a branch inside the production chain and select it at runtime using a random number generator

Question 10

A GenAI team at a retail company has built a RAG chatbot on Databricks that answers customer questions from a product catalog stored in Delta Lake. The catalog is updated nightly, and the team wants the chatbot to reflect those updates without manual intervention. They are deciding how to keep the Vector Search index synchronized. Which approach should they use?

Accepted Answer

Configure a Vector Search Delta Sync index on the catalog table so the index is refreshed automatically when the source table changes.. A Delta Sync index is the Databricks Vector Search construct that keeps an index aligned with a Delta table, including incremental changes. Because the catalog is a Delta table that changes nightly, configuring Delta Sync lets the chatbot retrieve current product information without a custom rebuild job. The other choices either fail to update the index or introduce unnecessary full recomputation.

Answer

Use a Databricks SQL scheduled query to copy the catalog table into a Parquet file and point the retriever at that file each night.

Answer

Create a standard Vector Search index and schedule a daily notebook that deletes the index and recreates it from the catalog table.

Answer

Enable Change Data Feed on the catalog table and rely on it to update the Vector Search index automatically.

Question 11

An engineer is designing a Databricks RAG application where the same retrieved context must be reused across several prompt variants during evaluation. They want to reduce token cost and improve consistency between variants. Which design choice best supports this?

Accepted Answer

Compute the retrieval step once, store the retrieved documents in a Delta table keyed by query, and have each prompt variant read from that table.. Separating retrieval from generation and persisting the retrieved context lets several prompt variants share one retrieval result. This lowers cost because embeddings and Vector Search are invoked once, and it improves consistency because all variants see the same evidence. The alternatives change generation or chunking behavior but do not eliminate redundant retrieval.

Answer

Use a higher temperature during evaluation so the model explores more responses and the best one can be selected.

Answer

Configure the model serving endpoint to cache identical requests so repeated prompts with the same context are served from cache.

Answer

Increase the chunk size in the Vector Search index so that each retrieval returns more context, reducing the need to call the retriever multiple times.

Question 12

A team is deploying a GenAI application to a Databricks Model Serving endpoint and wants to ensure the deployment can be rolled back quickly and that traffic is shifted safely during updates. Which TWO practices should the team follow? (Choose two.)

Accepted Answer

Configure the endpoint with multiple served entities and use traffic splitting to gradually shift traffic to the new model version.. Defining endpoints as code with Databricks Asset Bundles enables repeatable deployments and fast rollback by redeploying a prior version. Configuring multiple served entities with traffic splitting allows gradual rollout to the new version, so issues can be detected early and traffic reverted quickly, together providing safe updates and rollback.

Answer

Disable inference tables on the endpoint to reduce logging overhead during the update.

Answer

Deploy the new version by deleting the existing endpoint and creating a fresh one with the same name.

Answer

Set the endpoint's scale-to-zero behavior to always on to avoid cold starts during the update.

Question 13

A company is preparing a generative AI application for production and must demonstrate that model inputs and outputs are traceable and that access to sensitive prompt data is controlled. Which TWO Unity Catalog capabilities should the team rely on to meet these governance objectives? (Choose two.)

Accepted Answer

Fine-grained privileges on catalogs, schemas, tables, and models. Meeting the governance objectives requires both an enforcement mechanism and an evidence trail. Fine-grained privileges on the relevant catalogs, schemas, tables, and models enforce least-privilege access to sensitive prompt data, while Unity Catalog audit logs record which principals accessed those governed objects, providing the traceability that production compliance reviews demand.

Answer

Embedding caching to reduce inference latency

Answer

Vector index partitioning by document topic

Answer

Automatic hyperparameter tuning of registered models

Question 14

A GenAI engineer registers a fine-tuned model in Unity Catalog and wants a downstream application to call it through the Databricks Model Serving endpoint without embedding a long-lived personal access token in the application. The application runs on Azure Databricks and must authenticate as its own identity, and the security team requires that credentials be short-lived and automatically rotated. Which authentication approach should the engineer implement?

Accepted Answer

Use OAuth machine-to-machine authentication with the service principal's client ID and client secret to obtain a short-lived token. OAuth machine-to-machine authentication lets an application present a service principal's client ID and secret to the Databricks OAuth token endpoint and receive an access token with a limited lifetime. The application has its own Unity Catalog identity, permissions can be granted to that principal on the registered model and serving endpoint, and token refresh happens automatically. This is the supported pattern for non-interactive workloads that must avoid long-lived secrets.

Answer

Create a Databricks service principal, generate a personal access token for it, and store that token in the application's environment variables

Answer

Configure the application to use the workspace user's username and password for basic authentication

Answer

Embed a Microsoft Entra ID managed identity token directly in the application source code

Question 15

A GenAI engineer is building a RAG chatbot backed by a Databricks Vector Search index that is synced from a Delta table with Change Data Feed enabled. Source documents are frequently updated, and some are deleted. The engineer wants the index to reflect deletions and updates automatically with minimal operational overhead. Which configuration should be used?

Accepted Answer

Create a Delta Sync index with a pipeline_type of CONTINUOUS, pointing at the source Delta table, and let Databricks manage synchronization.. Automatic reflection of inserts, updates, and deletes from a Delta source into a Vector Search index is achieved with a Delta Sync index running in CONTINUOUS pipeline mode, which consumes the source table's Change Data Feed. Triggered refreshes, Direct Vector Access, and index-side Change Data Feed all either add manual work, introduce staleness, or rely on unsupported behavior.

Answer

Create a Direct Vector Access index and write a Structured Streaming job that upserts embeddings into it whenever the source table changes.

Answer

Create a Delta Sync index and enable Change Data Feed on the index itself so that deletions flow back into the source Delta table.

Answer

Create a Delta Sync index with a pipeline_type of TRIGGERED and schedule a Databricks Job to run a full refresh every 15 minutes.