EnglishDeutschFrançaisEspañolPortuguês

Retired

Snowflake · SF-GAI · Intermediate

SnowPro Specialty: Gen AI (GES-C02)

Validates specialized knowledge and best practices for leveraging Gen AI methodologies in Snowflake including Cortex LLM functions and AI governance.

This certification has been discontinued.

55Questions
85minTime Limit
750/ 1000Pass Score

Overview

About the exam

The SnowPro Specialty: Gen AI Certification (GES-C02) validates specialized knowledge of generative AI on Snowflake: the Cortex suite of AI features and functions, including Cortex Analyst, Cortex Search and Cortex Code; Gen AI governance across infrastructure, data and cost; document processing; and building and fine-tuning open-source models with Snowpark Container Services and the Snowflake Model Registry. It tests the ability to meet customer use cases natively on the platform.

This specialty certification is for AI and ML engineers, data scientists, data engineers and data application developers with one or more years of Gen AI experience with Snowflake in an enterprise environment. Proficiency writing code in Python helps, and previous data engineering and SQL knowledge is assumed.

Exam domains

What's on the exam

Exam blueprint: SF-GAI

Snowflake for Gen AI Overview18%

Define Snowflake's Gen AI principles and outline Gen AI capabilities across the platform, including Snowflake Intelligence and Cortex Code.

≈ 9 h
Snowflake Gen AI Functions38%

Apply AI functions in Snowflake, perform data analysis for a given use case, build or interact with interfaces, apply Cortex functions, and run third-party models.

≈ 19 h
Snowflake Gen AI Governance29%

Set up model access controls, grant and revoke role-based access, and manage, monitor and optimise Gen AI workloads.

≈ 14 h
Snowflake Document Processing15%

Use document parsing functions (AI_PARSE_DOCUMENT, AI_EXTRACT), prepare and manage documents, build automated pipelines with Streams and Tasks, and troubleshoot extraction.

≈ 8 h

Format

Exam format and question types

The exam consists of 55 questions in 85 minutes. Question types are multiple select and multiple choice. Questions cover four domains: Snowflake Gen AI Functions (38%), Snowflake Gen AI Governance (29%), Snowflake for Gen AI Overview (18%), and Snowflake Document Processing (15%). A passing score is 750 on a scale of 0 to 1000. At roughly 1.5 minutes per question, pacing is tight for scenario-based AI questions.

Question types: SF-GAI

Multiple Choice70%

Pick the single best answer from four or five options — the exam's bread and butter.

Multiple Response30%

More than one answer is correct and you need all of them; the question tells you how many to pick.

Try it now

Try five SF-GAI questions

Five questions straight from our SnowPro Specialty: Gen AI (GES-C02) pool. Answer one — Alex explains the why.

Snowflake Gen AI Governance1 / 5

By default, the USE AI FUNCTIONS privilege is granted to which role?

AlexFull explanation from Alex

PUBLIC USE AI FUNCTIONS: account-level privilege. Revokable by ACCOUNTADMIN only. Users need BOTH this AND CORTEX_USER or AI_FUNCTIONS_USER database role. Does not apply inside Snowflake native applications. SNOWFLAKE.CORTEX_USER: database role in SNOWFLAKE database. Covers all Covered AI features (Analyst, Search, Fine-tuning, Agent). Cannot be granted directly to users — must be granted to roles.

Sourcedocs.snowflake.com

Watch out

Common pitfalls

Candidates who have used general-purpose AI tools but haven't built with Snowflake Cortex specifically often struggle with questions about COMPLETE, EMBED, and SEARCH functions and their Snowflake-native integration patterns.

Watch list: SF-GAI

  1. 01Cortex COMPLETE vs External

    Not understanding when to use Cortex COMPLETE (native, serverless) versus calling external LLM APIs and the trade-offs of each approach leads to architecture question errors.

  2. 02RAG Pipeline Design

    Confusing embedding generation, vector storage, similarity search, and context injection steps in a RAG pipeline causes retrieval-augmented generation question failures.

  3. 03Vector Data Types

    Not understanding Snowflake VECTOR data type, distance metrics (cosine, inner product, L2), and how vector search integrates with Cortex leads to embedding question mistakes.

  4. 04Fine-Tuning vs Prompting

    Not knowing when fine-tuning is appropriate versus few-shot prompting or RAG — and the cost and data requirements of each — leads to wrong solution design answers.

Pass-IT trains you on exactly these weak spots — adaptive & spaced →

Related certifications

Related certifications

All Snowflake certifications