IBM AI Engineering Practice Test

A vector index is correct because The correct answer is a vector index because IBM uses it to store and retrieve document passages for retrieval-augmented generation patterns. The cited source, Creating a vector index programmatically, supports this answer for the Retrieval and Search scenario rather than the adjacent distractors.

Prompt evaluations expect variable-based inputs to map test data correctly is correct because The correct answer is that prompt evaluations expect variable-based inputs because the workflow maps prompt variables to test-data columns. The cited source, Evaluating prompt templates in projects, supports this answer for the Prompt Engineering and Evaluation scenario rather than the adjacent distractors.

Use the built-in Spark SQL `concat` or `concat_ws` functions inside a `select` or `withColumn` transformation is correct because To maintain scalable parallel execution in Apache Spark, computations must occur across the distributed worker nodes. The cited source, Spark SQL, DataFrames and Datasets Guide, supports this answer for the Apache Spark & Big Data for AI scenario rather than the adjacent distractors.

IBM Cloud IAM roles enforce strict access control. In a watsonx project, the 'Viewer' role grants read-only permission. Viewers can open and look at assets (like notebooks, prompt templates, or connection metadata) but cannot create, edit, or delete them. To modify and save a prompt template, the user must have the 'Editor' or 'Admin' role.

The correct answer is that AutoAI automates pipeline generation but still depends on informed experiment setup by the engineer. IBM's workflow still requires the user to define the model goal, training data, and relevant configuration choices.

For high-compliance environments, combining RAG (using a vector index for grounding) with automated governance is essential. By enabling prompt-template evaluation and tracking in watsonx.governance, the organization can automatically log every query-response event, monitor performance (using metrics like groundedness and safety), and generate audit-ready factsheets.

The correct answer is to promote or import the asset into the deployment space because spaces and projects are separate containers. IBM's deployment documentation makes that transfer step explicit before production deployment.

SHAP is mathematically rooted in cooperative game theory, calculating Shapley values to guarantee properties like local accuracy, consistency, and missingness, ensuring a fair distribution of feature attributions. LIME, on the other hand, is a local surrogate heuristic that perturbs inputs to build a linear approximation without the same theoretical guarantees.

project_id or space_id is correct because Every foundation model execution or inference call inside watsonx must be associated with a workspace container. The cited source, watsonx.ai Python SDK, supports this answer for the watsonx.ai Runtime SDK scenario rather than the adjacent distractors.

Vector databases compare high-dimensional embeddings using geometric metrics. The most common metrics are Cosine Similarity (which measures the angle between vectors, capturing direction or topic) and Euclidean (L2) Distance (which measures straight-line distance). These metrics determine how close two semantic representations are in the vector space.