I85PREP
⌘ K
watsonx Data Scientist - Associate study materials
watsonx Data Scientist - Associate study resources

watsonx Data Scientist - Associate practice questions

One question at a time. Build your watsonx Data Scientist - Associate understanding.

First 50 of 200 questions. Each one lists the choices, the correct answer, and an explanation. The checker at the top of this page saves progress in this browser when it is available.

Continue in the practice tool above to load more questions, 50 at a time.

Question 1watsonx Platform Fundamentals and Architecture

In the watsonx platform, which component is the governed data store that holds curated, cataloged data for analysis and model training?

  1. watsonx.governance
  2. watsonx.data Correct answer.
  3. watsonx.ai
  4. watsonx Orchestrate

Explanation

watsonx.data is the open lakehouse layer that stores and queries governed data across object storage and multiple engines, making it the platform's curated data foundation for analytics and model development. The other named services address model building, risk governance, and task automation rather than serving as the governed data store.

Question 2watsonx Platform Fundamentals and Architecture

A team wants a single place to discover, tune, and deploy both IBM-provided and open-source foundation models. Which watsonx capability should they use?

  1. watsonx.ai Correct answer.
  2. watsonx Code Assistant
  3. watsonx.data
  4. watsonx.governance

Explanation

watsonx.ai provides the model catalog, prompt tooling, tuning environments, and deployment runtimes for foundation models, so it is the correct home for discovering and operationalizing models. The other options focus on data storage, governance, or code assistance.

Question 3watsonx Platform Fundamentals and Architecture

Which statement best describes the role of a project in watsonx.ai?

  1. It is a governance record that must be approved before any model can be trained.
  2. It is a billing account that aggregates usage charges across all platform services.
  3. It is a runtime cluster that replaces the need for a deployment space.
  4. It is a collaborative workspace that groups data assets, notebooks, models, and deployments for a team. Correct answer.

Explanation

A project is the collaborative working boundary where team members assemble data assets, notebooks, prompts, and trained models. Deployment spaces are separate artifacts used to promote and serve assets, and billing or mandatory approval records are handled elsewhere.

Question 4watsonx Platform Fundamentals and Architecture

What is the primary purpose of a deployment space in watsonx.ai?

  1. To author prompt templates collaboratively
  2. To store raw training data before it is cataloged
  3. To host promoted assets and serve them as online or batch deployments Correct answer.
  4. To record model risk assessments for auditors

Explanation

A deployment space is the promotion target where assets from a project are placed and then exposed through online endpoints or batch jobs. Authoring prompts and storing raw data happen in projects or the lakehouse, while risk assessments belong to governance tooling.

Question 5watsonx Platform Fundamentals and Architecture

A data scientist has a trained model in a watsonx.ai project and needs to expose it as a REST endpoint for an application. What is the required next step?

  1. Export the notebook and run it on the application server.
  2. Publish the model to a deployment space and create an online deployment there. Correct answer.
  3. Register the model in watsonx.data and query it with SQL.
  4. Convert the project to a catalog and grant the application access.

Explanation

Serving a model through an endpoint requires promoting it to a deployment space, where an online deployment generates a scoring endpoint. Projects are authoring environments, and neither catalogs nor the lakehouse SQL engine provide model scoring endpoints.

Question 6watsonx Platform Fundamentals and Architecture

Which watsonx component provides factsheets, model inventory tracking, and monitoring of model fairness and drift?

  1. watsonx Assistant
  2. watsonx.governance Correct answer.
  3. watsonx.ai
  4. watsonx.data

Explanation

watsonx.governance centralizes AI asset inventory, automated factsheets, and ongoing evaluation of fairness, drift, and risk throughout the lifecycle. The other services cover data, model building, or conversational assistants.

Question 7watsonx Platform Fundamentals and Architecture

How does the watsonx platform architecture typically separate the control plane from the compute or runtime plane?

  1. The control plane runs inside the customer's browser, while the runtime plane is always on-premises.
  2. The control plane executes queries, while the runtime plane handles user authentication.
  3. The control plane stores all customer data, while the runtime plane only displays dashboards.
  4. The control plane manages provisioning, identity, and service configuration, while the runtime plane executes workloads such as model inference and SQL queries. Correct answer.

Explanation

In a cloud service architecture, the control plane handles orchestration concerns such as provisioning, identity, and configuration, whereas the runtime plane performs the actual workload execution like inference or query processing. This separation supports scalability and independent operations.

Question 8watsonx Platform Fundamentals and Architecture

Which IBM Cloud service is commonly used as the identity and access management foundation for a watsonx deployment?

  1. IBM Cloud Object Storage
  2. IBM Cloud Databases
  3. IBM Cloud IAM Correct answer.
  4. IBM Event Streams

Explanation

IBM Cloud IAM provides authentication, service IDs, API keys, and role-based access policies that govern watsonx resources. Object Storage, Event Streams, and managed databases are workload services rather than the identity layer.

Question 9watsonx Platform Fundamentals and Architecture

A user can open a watsonx.ai project but cannot create a deployment in the associated space. Which access mechanism most directly explains this?

  1. The project's data assets are encrypted with a customer-managed key.
  2. The model was trained with a foundation model that lacks a license.
  3. The user's API key has expired for all services.
  4. IAM policies grant different roles on the project and on the deployment space resource. Correct answer.

Explanation

Access in watsonx is granted per resource through IAM roles, so a user may hold an editor role on a project yet only a viewer role on the space. Expired keys or encryption settings would block broader access rather than selectively prevent deployment creation.

Question 10watsonx Platform Fundamentals and Architecture

What does the term foundation model mean in the context of watsonx.ai?

  1. A rules-based engine that encodes regulatory compliance logic
  2. A small classical model used exclusively for tabular regression
  3. A model trained only on a single organization's labeled dataset from scratch
  4. A large pretrained model that can be adapted to many downstream tasks through prompting or tuning Correct answer.

Explanation

Foundation models are pretrained on broad data at scale and then adapted to specific tasks via prompting, retrieval, or fine-tuning. Models built only for one labeled dataset or small tabular regressors are not foundation models.

Question 11watsonx Platform Fundamentals and Architecture

A team wants to ground a foundation model's answers in their own documents without retraining the model. Which watsonx.ai approach fits best?

  1. Increasing the model's temperature parameter
  2. Deploying the model as a batch job instead of online
  3. Retrieval-augmented generation using a vector index of the documents Correct answer.
  4. Full fine-tuning of every model weight on the document corpus

Explanation

Retrieval-augmented generation retrieves relevant passages from an indexed corpus and supplies them as context, grounding answers without changing model weights. Fine-tuning is heavier, and temperature or deployment mode do not inject document knowledge.

Question 12watsonx Platform Fundamentals and Architecture

Which statement accurately describes watsonx.data's open lakehouse design?

  1. It queries data in open table formats across object storage using multiple engines such as Presto and Spark. Correct answer.
  2. It requires all data to be loaded into a proprietary columnar warehouse before querying.
  3. It replaces object storage with block storage attached to each compute node.
  4. It supports only CSV files and cannot use open table formats.

Explanation

watsonx.data separates storage from compute and lets multiple query engines work over open table formats held in object storage. That design avoids proprietary lock-in and does not require proprietary warehouse loading or block storage.

Question 13watsonx Platform Fundamentals and Architecture

A user needs to run interactive SQL against tables registered in watsonx.data. Which engine is designed for that low-latency query use case?

  1. Watson OpenScale
  2. IBM watsonx.governance
  3. Presto Correct answer.
  4. IBM Cloud IAM

Explanation

Presto is a distributed SQL query engine included in watsonx.data for interactive, low-latency analytics over lakehouse tables. The other options are governance or identity services, not query engines.

Question 14watsonx Platform Fundamentals and Architecture

Which scenario best justifies using a deployment space instead of keeping everything inside a project?

  1. An analyst wants to browse a catalog of data assets.
  2. A reviewer wants to read a model's automated factsheet.
  3. An application team needs a stable scoring endpoint that persists independently of project notebook changes. Correct answer.
  4. A data scientist wants to edit a notebook with teammates in real time.

Explanation

Deployment spaces provide a stable serving boundary with endpoints and access control that is decoupled from ongoing project authoring. Collaborative notebook editing, catalog browsing, and factsheet review are project or governance activities.

Question 15watsonx Platform Fundamentals and Architecture

Which capability is provided by the watsonx.ai Prompt Lab?

  1. Enforcing data residency rules for stored tables
  2. Scheduling batch scoring jobs across deployment spaces
  3. Automatically retraining a foundation model on every prompt submitted
  4. Experimenting with prompt templates and model parameters against sample inputs before integrating a model Correct answer.

Explanation

Prompt Lab lets practitioners draft and compare prompts, adjust parameters such as temperature, and inspect outputs before wiring a model into an application. It does not retrain models, enforce residency policy, or schedule batch jobs.

Question 16watsonx Platform Fundamentals and Architecture

What is the purpose of the watsonx.ai model catalog?

  1. To store the organization's proprietary training datasets
  2. To track deployment endpoint latency metrics
  3. To list available foundation models with their providers, capabilities, and suitability for tasks Correct answer.
  4. To define IAM roles for project members

Explanation

The catalog documents which foundation models are available, who provides them, and what tasks they suit, helping teams choose appropriately. Datasets, latency metrics, and IAM roles are managed by other platform features.

Question 17watsonx Platform Fundamentals and Architecture

A model in production shows declining accuracy as customer behavior shifts over time. Which governance practice addresses this in watsonx?

  1. Disabling the model's deployment endpoint permanently
  2. Increasing the batch size of the training job
  3. Continuous model monitoring that detects drift and triggers evaluation or retraining workflows Correct answer.
  4. Adding more IAM editors to the deployment space

Explanation

Drift monitoring compares production behavior against expectations and raises alerts that feed evaluation and retraining. Removing the endpoint, changing batch size, or broadening access does not detect or correct model degradation.

Question 18watsonx Platform Fundamentals and Architecture

Which pairing of watsonx service and primary responsibility is correct?

  1. watsonx.data for model serving, watsonx.ai for data storage
  2. watsonx.governance for model tuning, watsonx.data for risk factsheets
  3. watsonx.data for lakehouse storage and query, watsonx.ai for model development and serving Correct answer.
  4. watsonx.ai for identity management, watsonx.governance for SQL query

Explanation

The platform divides responsibilities: watsonx.data handles governed lakehouse storage and query, while watsonx.ai handles model development, tuning, and serving. The other pairings invert or misassign these core roles.

Question 19watsonx Platform Fundamentals and Architecture

What advantage does watsonx.data's separation of storage and compute provide?

  1. It forces all workloads onto a single engine to simplify operations.
  2. It removes the need for access control on tables.
  3. Teams can scale query engines independently of stored data volume and run multiple engines over the same data. Correct answer.
  4. It guarantees that all queries complete without any resource provisioning.

Explanation

Decoupling storage from compute lets organizations attach or resize engines on demand and use different engines against shared object-stored data. It does not eliminate provisioning, access control, or the option of multiple engines.

Question 20watsonx Platform Fundamentals and Architecture

Which practice best supports reproducibility of a watsonx.ai experiment across team members?

  1. Storing prompts in personal email drafts
  2. Recording the model, prompt, parameters, and data asset versions in the project so others can rerun them Correct answer.
  3. Sharing only the final deployment endpoint URL
  4. Relying on each member's local notebook copy without versioning

Explanation

Reproducibility requires capturing the inputs that produced a result, including model choice, prompt text, parameters, and data versions, within the shared project. Local copies, endpoint URLs, or informal storage lose the context needed to rerun work.

Question 21watsonx Platform Fundamentals and Architecture

Which statement about watsonx.governance factsheets is accurate?

  1. Factsheets consolidate model details such as purpose, training data, performance, and risk considerations for stakeholders. Correct answer.
  2. Factsheets replace the need for model monitoring once published.
  3. Factsheets are automatically generated SQL views over lakehouse tables.
  4. Factsheets are only available for classical models, not foundation models.

Explanation

Factsheets document a model's purpose, data, evaluation results, and risk factors in a shareable form for reviewers and auditors. They are governance documents, not SQL views, and they complement rather than replace ongoing monitoring.

Question 22watsonx Platform Fundamentals and Architecture

A team must prove to regulators that a deployed model was evaluated for bias before release. Which watsonx capability most directly produces this evidence?

  1. The deployment space's endpoint URL
  2. The lakehouse's Presto query history
  3. Governance evaluations that record fairness metrics and attach them to the model's factsheet Correct answer.
  4. The Prompt Lab's temperature slider

Explanation

Governance evaluations compute fairness and quality metrics and preserve them as auditable evidence linked to the model. Prompt parameters, query history, and endpoint addresses do not document bias assessment.

Question 23watsonx Platform Fundamentals and Architecture

Which watsonx.ai tuning approach updates a small set of additional parameters while keeping the base model weights frozen?

  1. Full fine-tuning of all model weights
  2. Prompt engineering without any training
  3. Parameter-efficient tuning such as prompt tuning or low-rank adaptation Correct answer.
  4. Retrieval-augmented generation

Explanation

Parameter-efficient methods train a small adapter or prompt embedding while leaving the base model frozen, reducing cost and storage. Full fine-tuning changes all weights, and prompting or retrieval do not train parameters.

Question 24watsonx Platform Fundamentals and Architecture

What is the role of a vector index in a watsonx.ai retrieval-augmented generation pipeline?

  1. It schedules batch scoring jobs on a timer
  2. It stores document embeddings so relevant passages can be retrieved by similarity for the prompt context Correct answer.
  3. It stores the model's weights for deployment
  4. It enforces IAM policies on the deployment space

Explanation

A vector index holds embeddings of document chunks and supports nearest-neighbor search so the most relevant passages can be supplied as context. Model weights, IAM policy, and job scheduling are handled by other components.

Question 25Data Preparation and Governance with watsonx.data and watsonx.governance

In watsonx.data, which component stores table and column metadata and serves as the catalog that query engines use to resolve schemas?

  1. The metastore service Correct answer.
  2. The Presto (Prestissimo) worker pool
  3. The object storage bucket
  4. The ingestion job scheduler

Explanation

watsonx.data uses a metastore service to hold catalog, schema, table, and column definitions so engines such as Presto and Spark can resolve names consistently. Object storage holds the data files, while compute engines only read metadata from the metastore.

Question 26Data Preparation and Governance with watsonx.data and watsonx.governance

A data engineer registers a Parquet dataset stored in an S3-compatible bucket as a watsonx.data table. Which statement describes the relationship between the metastore entry and the underlying files?

  1. The metastore holds the schema and file location while the rows remain in object storage Correct answer.
  2. The metastore compresses the files and rewrites them into its internal format
  3. The metastore deletes the source files after the first successful query
  4. The metastore copies all rows into its own relational tables

Explanation

Registering an external table creates metadata that points at the storage location; the data itself stays in the bucket. This separation lets multiple engines query the same files without duplication.

Question 27Data Preparation and Governance with watsonx.data and watsonx.governance

Which watsonx.data capability allows a single query to join a table in a Presto-managed catalog with a table in a Hive catalog without moving data first?

  1. Replication of both tables into a single bucket
  2. Schema versioning in the metastore
  3. Manual export of both tables to CSV
  4. Federated query across multiple registered catalogs Correct answer.

Explanation

watsonx.data exposes multiple catalogs to one engine, so federated queries can join across them in place. Copying data first is unnecessary and introduces latency and duplication.

Question 28Data Preparation and Governance with watsonx.data and watsonx.governance

What is the primary purpose of a governance policy in watsonx.governance when applied to a data asset used for model training?

  1. To define and enforce rules such as permitted use and required approvals Correct answer.
  2. To convert the asset into a different file format
  3. To schedule automatic retraining of the model
  4. To increase the storage quota of the asset

Explanation

Governance policies codify rules about how an asset may be used, who must approve access, and what controls apply. They are enforcement and accountability mechanisms rather than storage or transformation features.

Question 29Data Preparation and Governance with watsonx.data and watsonx.governance

Which artifact in watsonx.governance records the lineage, ownership, and quality attributes of a data asset so stewards can review it?

  1. A notebook kernel specification
  2. A model deployment endpoint
  3. A query execution plan
  4. A data asset entry in the governance catalog Correct answer.

Explanation

Governance catalogs hold asset entries enriched with ownership, lineage, and quality metadata for steward review. Deployment endpoints and execution plans are runtime artifacts without stewardship metadata.

Question 30Data Preparation and Governance with watsonx.data and watsonx.governance

A steward needs to mask customer email addresses for analysts who query a watsonx.data table. Which approach applies governance at query time without altering the stored files?

  1. Deleting the email column from the table
  2. Granting analysts write access to the bucket
  3. Creating a governed view that applies a masking expression Correct answer.
  4. Rewriting every Parquet file with hashed emails

Explanation

A governed view applies masking logic at query time, so the underlying files remain unchanged while analysts see redacted values. Rewriting files changes the source of truth and is harder to reverse.

Question 31Data Preparation and Governance with watsonx.data and watsonx.governance

An ingestion job loads CSV files into a watsonx.data managed table, and some rows contain a date string that does not match the target DATE type. What is the most appropriate handling strategy during ingestion?

  1. Route failing rows to a quarantine location and log the rejection reason Correct answer.
  2. Drop the entire batch and disable ingestion
  3. Silently coerce invalid values to the current date
  4. Change the column type to STRING permanently for all future loads

Explanation

Quarantining rejected rows preserves them for remediation while letting valid rows load, and logging makes the failure auditable. Silent coercion corrupts data, and dropping whole batches hides the real defect.

Question 32Data Preparation and Governance with watsonx.data and watsonx.governance

A team wants watsonx.data queries to read only the columns and partitions they request from a large Parquet table. Which table design most directly supports that behavior?

  1. Partitioned Parquet files with column statistics in the metastore Correct answer.
  2. An external table pointing at a compressed archive of mixed files
  3. A view that concatenates every column into one string
  4. A single uncompressed CSV file in the bucket root

Explanation

Partitioned Parquet enables partition pruning and columnar reads so engines touch only needed data. Column statistics further help the optimizer skip irrelevant files.

Question 33Data Preparation and Governance with watsonx.data and watsonx.governance

A governance officer must demonstrate that a training dataset was approved by the data owner before use. Which watsonx.governance record best provides this evidence?

  1. The model's inference latency metrics
  2. The query runtime statistics of the training job
  3. The storage class configured on the bucket
  4. An approval record attached to the data asset's governance workflow Correct answer.

Explanation

Approval records captured in the governance workflow provide auditable evidence of sign-off by the accountable owner. Runtime and storage metrics say nothing about authorization.

Question 34Data Preparation and Governance with watsonx.data and watsonx.governance

A data steward is told that a column previously classified as public now contains personal data. What is the correct governance action in watsonx.governance?

  1. Move the asset to a bucket with public read access
  2. Update the asset's classification and propagate the associated policy controls Correct answer.
  3. Disable lineage tracking for the asset
  4. Delete the asset entry and recreate it under a new name

Explanation

Reclassification should update the asset's sensitivity label so the linked policies, such as masking or access restrictions, take effect. Deleting or hiding the asset breaks lineage and audit history.

Question 35Data Preparation and Governance with watsonx.data and watsonx.governance

Which statement best explains why lineage capture in watsonx.governance matters when a data quality defect is discovered in a source table?

  1. It increases the compression ratio of the source files
  2. It identifies downstream assets and models that consumed the affected data Correct answer.
  3. It automatically repairs the defective rows
  4. It replaces the need for access control policies

Explanation

Lineage maps upstream and downstream dependencies, so teams can find every asset and model impacted by a defect. It supports impact analysis but does not repair data itself.

Question 36Data Preparation and Governance with watsonx.data and watsonx.governance

A compliance rule requires that datasets used for a model be retained for seven years and that deletions be documented. Which combination of watsonx.data and watsonx.governance features addresses this requirement?

  1. Presto worker autoscaling plus query caching
  2. Object storage lifecycle rules plus governance retention and deletion audit records Correct answer.
  3. Parquet compression plus column statistics
  4. Metastore backups plus notebook versioning

Explanation

Lifecycle rules on storage enforce retention periods, while governance records document retention policy and deletion events for audit. Scaling and compression settings do not create retention evidence.

Question 37Data Preparation and Governance with watsonx.data and watsonx.governance

When a data engineer adds a new column to a Parquet dataset registered as a watsonx.data table, what must happen for queries to see the column without reloading data?

  1. The metastore schema must be refreshed or altered to include the new column Correct answer.
  2. All existing files must be rewritten with the old schema
  3. The bucket must be recreated with a new name
  4. The query engine must be reinstalled

Explanation

Because the metastore holds the schema, refreshing or altering it lets engines read the new column from the existing files. No data movement or engine reinstallation is required.

Question 38Data Preparation and Governance with watsonx.data and watsonx.governance

A reviewer asks why a governed data asset shows both a technical owner and a business owner. Which explanation is most accurate?

  1. The two roles are duplicates and one can always be removed
  2. Technical owners handle pipeline reliability while business owners approve usage and interpret policy Correct answer.
  3. The business owner runs the ingestion jobs and the technical owner approves usage
  4. Both roles exist only to satisfy a licensing requirement

Explanation

Separating technical and business ownership assigns operational accountability and policy accountability to the right people. This division makes approvals and incident handling clearer.

Question 39Data Preparation and Governance with watsonx.data and watsonx.governance

A team must prove that a model's training data came only from approved sources. Which governance evidence is most direct?

  1. The engine version used during training
  2. The model card's intended-use section alone
  3. The number of partitions in the training table
  4. Lineage from the training dataset to its registered source assets with approval status Correct answer.

Explanation

Lineage ties the training dataset back to registered sources, and each source's approval status shows whether it was authorized. Partition counts and engine versions do not establish provenance.

Question 40Data Preparation and Governance with watsonx.data and watsonx.governance

An analyst reports that a governed view returns fewer rows than the base table. Which explanation is most consistent with governance design?

  1. The view applies a row-level filter that restricts records to those the analyst is entitled to see Correct answer.
  2. The object storage bucket lost durability
  3. The view silently corrupts data during reads
  4. The metastore deleted rows to save space

Explanation

Row-level filtering in a governed view restricts visible records according to entitlement, which legitimately reduces row counts. Storage durability and metastore behavior do not selectively hide rows.

Question 41Data Preparation and Governance with watsonx.data and watsonx.governance

Which practice best supports reproducible data preparation in watsonx.data when multiple engineers transform the same source table?

  1. Ad-hoc manual edits performed directly in a notebook with no saved script
  2. Granting every engineer direct write access to the source bucket
  3. Version-controlled transformation definitions executed through scheduled pipelines Correct answer.
  4. Emailing CSV extracts to each engineer

Explanation

Version-controlled transformations executed by pipelines make each change traceable and repeatable. Ad-hoc edits and emailed extracts create divergent, unauditable copies.

Question 42Data Preparation and Governance with watsonx.data and watsonx.governance

A governance team wants automated detection of sensitive data such as national identifiers in newly registered watsonx.data tables. Which capability should they configure?

  1. Query result caching in the engine
  2. Bucket versioning for object recovery
  3. Automated sensitive-data profiling that assigns classifications to detected columns Correct answer.
  4. Manual row counting during ingestion

Explanation

Automated profiling inspects column values and patterns to detect sensitive categories and assign classifications that drive policy. Row counts, caching, and versioning do not identify sensitive content.

Question 43Data Preparation and Governance with watsonx.data and watsonx.governance

A data engineer needs to enforce that only rows from approved regions appear in a shared reporting table without maintaining separate physical copies per region. Which design achieves this with the least duplication?

  1. One physical table per region, each loaded nightly
  2. A separate watsonx.data instance per region
  3. A single base table with a governed view that filters by region entitlement Correct answer.
  4. A CSV export per region stored in a shared bucket

Explanation

A single base table plus a filtering governed view enforces regional visibility while avoiding duplicate storage. Per-region copies multiply cost and drift risk.

Question 44Data Preparation and Governance with watsonx.data and watsonx.governance

A steward observes that a table's governance classification is 'Confidential' but the linked policy still permits broad analyst access. What is the most likely root cause?

  1. The metastore has run out of disk space
  2. The Parquet files use an unsupported compression codec
  3. The query engine lacks enough worker nodes
  4. The policy is not bound to the classification or the asset, so the label has no enforcement effect Correct answer.

Explanation

Classifications only enforce controls when policies are attached to them and to the asset. Without that binding, the label is descriptive metadata with no access effect.

Question 45Data Preparation and Governance with watsonx.data and watsonx.governance

Two catalogs in watsonx.data contain tables with the same name in the same schema. A federated query must reference each unambiguously. What is the correct approach?

  1. Fully qualify each reference with catalog, schema, and table name Correct answer.
  2. Rename both tables to identical names in a third catalog
  3. Disable one catalog before running the query
  4. Rely on the engine to pick the most recently created table

Explanation

Fully qualified names resolve ambiguity because the catalog component distinguishes the otherwise identical schema and table names. Relying on creation order is nondeterministic and unsafe.

Question 46Data Preparation and Governance with watsonx.data and watsonx.governance

An auditor asks how the team can show that a masking rule was applied consistently to a dataset over time. Which evidence combination is strongest?

  1. Screenshots of one query run and a verbal confirmation
  2. Versioned policy definitions plus lineage showing the governed view used by downstream consumers Correct answer.
  3. The storage billing report for the affected bucket
  4. The count of columns in the base table

Explanation

Versioned policies show what rule applied at each point, and lineage shows which consumers used the governed view. Together they demonstrate consistent, traceable enforcement.

Question 47Data Preparation and Governance with watsonx.data and watsonx.governance

A pipeline ingests JSON events whose schema varies by event type. The team wants queries to work without rejecting valid events that omit optional fields. Which ingestion approach is most appropriate?

  1. Store events with a flexible schema and expose a governed view that projects a stable set of fields Correct answer.
  2. Convert all events to fixed-width text before loading
  3. Load each event type into a separate watsonx.data instance
  4. Enforce a single rigid schema and reject every event missing any field

Explanation

Flexible storage accommodates varying event shapes, and a governed view presents a stable projection for consumers while applying policy. Rigid rejection or instance sprawl adds cost and data loss.

Question 48Data Preparation and Governance with watsonx.data and watsonx.governance

A governance lead must decide whether to block a dataset from model training because its quality score fell below threshold. Which factor is most relevant to that decision?

  1. The documented quality thresholds and the asset's measured quality results Correct answer.
  2. The engine's worker count during profiling
  3. The number of partitions in the dataset
  4. The compression codec used by the files

Explanation

The decision should compare measured quality results against documented thresholds tied to the asset's intended use. Physical layout and compute settings do not determine fitness for training.

Question 49Data Preparation and Governance with watsonx.data and watsonx.governance

In watsonx.data, a table is registered in a catalog with a schema and table name. A federated query engine must resolve that table uniquely. Which identifier form correctly and unambiguously references the table?

  1. schema.table
  2. table
  3. catalog.table
  4. catalog.schema.table Correct answer.

Explanation

A fully qualified three-part name in watsonx.data follows the pattern catalog.schema.table, so the catalog component disambiguates tables that share schema and table names across catalogs. Omitting any part leaves the engine to resolve it from context, which can select the wrong table or fail.

Question 50Data Preparation and Governance with watsonx.data and watsonx.governance

A data engineer loads Parquet files into a watsonx.data managed table and later wants to add a new column to the source files. The team still wants existing queries to keep working. Which approach best preserves query compatibility?

  1. Store the new column in a separate table and require joins for every query
  2. Add the new column with a defined default or nullable type so existing readers resolve missing values Correct answer.
  3. Rename the table so old queries fail fast and users adopt the new schema
  4. Overwrite all files with the new column populated and re-register the table from scratch

Explanation

Schema evolution in columnar formats works best when new columns are added as nullable or with defaults, allowing readers to return nulls for older files that lack the field. This keeps existing queries valid while exposing the new attribute to new queries.