GES-P02 SnowPro Practice Exam: Gen AI Practice Questions
Prepare for GES-P02 with more than an answer.
- Level
- Specialty (Practice Exam)
Domains covered on the exam 4
- Snowflake for Gen AI Overview18%
- Snowflake Gen AI Functions38%
- Snowflake Gen AI Governance29%
- Snowflake Document Processing15%
- 1
An architect is designing a Retrieval-Augmented Generation (RAG) system using Snowflake Cortex Search.
flowchart LR Doc[Documents] --> A[Step A] A --> B[Step B] B --> C[Cortex Search Service]To properly ingest text data so it can be searched semantically, which TWO processes must occur during the ingestion phase (Steps A and B)? (Select TWO)
Show answer details
Correct answer: A, B
To perform semantic search, long documents must be chunked into smaller, meaningful segments. These chunks are then converted into vector embeddings (mathematical representations of meaning) which are stored and indexed by the Cortex Search service.
To perform semantic search, long documents must be chunked into smaller, meaningful segments. These chunks are then converted into vector embeddings (mathematical representations of meaning) which are stored and indexed by the Cortex Search service.
- 2
A company is deploying Cortex Analyst to allow business users to ask natural language questions about their sales data. The LLM occasionally fabricates SQL columns that do not exist in the actual schema. What is the REQUIRED best practice configuration to ground Cortex Analyst and ensure it generates accurate Text-to-SQL queries?
Show answer details
Correct answer: C
Cortex Analyst relies on a Semantic Model—typically a YAML configuration file stored on an internal stage—to understand the relationships, synonyms, and exact schema structures of the underlying data. This grounds the Text-to-SQL generation and prevents hallucinations.
- 3
A data team previously built a pipeline that calls
SNOWFLAKE.CORTEX.COMPLETEinside a User-Defined Function (UDF) to determine if product reviews mention pricing complaints. This approach processes row-by-row and is experiencing performance bottlenecks.They want to migrate to a more efficient, vectorized approach using the expanded AISQL function suite designed specifically for this type of conditional extraction.
Which strategy achieves this optimization?
Show answer details
Correct answer: C
The
AI_FILTERfunction in the AISQL suite allows users to filter tables based on semantic meaning or conditions directly in standard SQL. It is optimized for batch operations over tables, replacing slow, row-by-row UDF implementations ofCOMPLETE.flowchart TD Old[Row-by-Row COMPLETE] -->|Slow| Bottleneck(Performance Issue) New[SELECT * FROM reviews WHERE AI_FILTER(text, 'mentions pricing')] -->|Vectorized| Fast(Optimized SQL) - 4
A financial services company is evaluating Snowflake Cortex AI against external LLM providers for processing highly sensitive customer transaction summaries. The Chief Information Security Officer (CISO) mandates that no customer data can leave the organization's governed security perimeter. Which characteristic of Snowflake Cortex AI natively satisfies this compliance requirement?
Show answer details
Correct answer: B
Snowflake Cortex AI provides fully managed, serverless LLM inference directly within the Snowflake Data Cloud. Because the models run inside Snowflake's secure perimeter, customer data is never transmitted to external third-party APIs, ensuring strict data-residency and compliance requirements are met natively.
- 5
A data engineering team wants to orchestrate a complex agentic workflow where an AI agent can autonomously query sales data, execute Python scripts, and summarize findings. Which Snowflake feature is specifically designed to orchestrate and manage these agentic tools and workflows?
Show answer details
Correct answer: D
Snowflake Intelligence is a platform designed to create and manage agentic AI workflows. It allows developers to build agents that can reason, use tools (like SQL execution or Python scripts), and orchestrate multi-step tasks across enterprise data.
- 6
A healthcare organization requires a highly customized, open-source medical LLM (e.g., MedLlama) to process patient records. They cannot use managed Cortex functions due to the need for custom model weights. Which TWO Snowflake features must be utilized together to host and serve this custom open-source model securely? (Select TWO)
Show answer details
Correct answer: C, D
To deploy custom or open-source models with specific weights, organizations use SPCS (which supports GPU compute pools) to host the model securely, and the Snowflake Model Registry to manage, version, and deploy the model artifacts.
flowchart LR MR[Model Registry] -->|Deploy| SPCS[Snowpark Container Services] SPCS -->|Inference| DB[(Snowflake Data)]To deploy custom or open-source models with specific weights, organizations use SPCS (which supports GPU compute pools) to host the model securely, and the Snowflake Model Registry to manage, version, and deploy the model artifacts.
flowchart LR MR[Model Registry] -->|Deploy| SPCS[Snowpark Container Services] SPCS -->|Inference| DB[(Snowflake Data)]
