SNOWPRO SPECIALTY:
GEN AI
EXAM STUDY GUIDE
Last Updated: August 22, 2025
TABLE OF CONTENTS:
SNOWPROⓇ SPECIALTY: GEN AI STUDY GUIDE OVERVIEW........................ 3
RECOMMENDATIONS FOR USING THIS GUIDE................................................... 3
SNOWPROⓇ SPECIALTY: GEN AI EXAM CONTENT AND FORMAT............... 4
CERTIFICATION OVERVIEW.................................................................................... 4
SNOWPROⓇ SPECIALTY: GEN AI PREREQUISITE.............................................. 5
STEPS TO SUCCESS....................................................................................................5
EXAM DOMAINS AND WEIGHTINGS....................................................................... 5
Domain 1.0: Snowflake for Gen AI Overview.................................................................... 6
Domain 2.0: Snowflake Gen AI & LLM Functions............................................................ 8
Domain 3.0: Snowflake Gen AI Governance.................................................................... 10
Domain 4.0: Snowflake Document AI...............................................................................12
SNOWPROⓇ SPECIALTY: GEN AI SAMPLE QUESTIONS................................. 13
MAINTAINING YOUR CERTIFICATION................................................................ 15
Page 2
SNOWPROⓇ SPECIALTY: GEN AI STUDY GUIDE OVERVIEW
This is a self-paced guide that will outline the Snowflake domains, objectives, and topics that
will be covered on the SnowPro Specialty: Gen AI Certification Exam. Use of this study
guide does not guarantee certification success.
Holding either the SnowPro Associate: Platform or the SnowPro Core Certification in good
standing are prerequisite options for taking the Specialty: Gen AI Certification Exam.
For an overview and more information on all of the available SnowPro Certifications, please
navigate here.
RECOMMENDATIONS FOR USING THIS GUIDE
This guide outlines the Snowflake topics and subtopics covered on the exam. Following the
topics will be additional resources consisting of links to documentation, blogs, and/or exercises
that are designed to support candidates’ understanding of Gen AI on Snowflake.
Estimated length of study time required to complete the guide: 10 – 13 hours
Some links may have more value for exam preparation than others, depending on the candidate’s
level of experience. The same amount of time should not be spent on each link. Some links may
appear in more than one content domain.
Page 3
SNOWPROⓇ SPECIALTY: GEN AI EXAM CONTENT AND FORMAT
CERTIFICATION OVERVIEW
The SnowPro Specialty: Gen AI Certification Exam tests specialized knowledge, skills, and
best practices used to leverage Gen AI methodologies in Snowflake including key concepts,
features, and programming constructs. The exam will assess skills through scenario-based
questions, interactive questions, and real-world examples.
This certification will test the candidate’s ability to:
● Define and implement Snowflake Gen AI principles, capabilities, and best practices
related to infrastructure, data governance, and cost governance
● Leverage Snowflake Cortex AI features, Large Language Models (LLMs), and offerings
to support customer use cases (for example, Cortex Analyst, Cortex Search, Cortex
Fine-tuning, Snowflake Copilot)
● Build open-source models with Snowpark Container Services and Snowflake Model
Registry (for example, Hugging Face)
● Use Document AI to train and troubleshoot models specific to authentic use cases
The candidate is expected to have knowledge of:
● Snowflake Cortex suite of Gen AI features and their underlying models
● Retrieval Augmented Generation (RAG) applications leveraging LLMs
Target Audience:
Candidates with one or more years of Gen AI experience with Snowflake, in an enterprise
environment. In addition, successful candidates may have advanced proficiency writing code in
Python. Previous data engineering and SQL knowledge is assumed.
This exam is designed for:
● AI or ML Engineers
● Data Scientists
● Data Engineers
● Data Application Developers
● Data Analysts with programming experience
Page 4
SNOWPROⓇ SPECIALTY: GEN AI PREREQUISITE
To be eligible to take this exam candidates must hold an active SnowPro Associate: Platform or
SnowPro Core Certification.
STEPS TO SUCCESS
1. Review this Gen AI Exam Study Guide
2. Attend Snowflake’s Instructor-Led Snowflake Gen AI Training
3. Get hands-on experience with our Snowflake x Gen AI: A Cool Collab and Snowflake X
Gen AI: LLM Functions
4. Review and study applicable white papers and documentation
5. Get hands-on practical experience with relevant business requirements using Snowflake
6. Attend free Snowflake Webinars
7. Attend Snowflake Virtual Hands-on Labs to gain practical experience
8. Practice with sample questions here
9. Schedule the exam
10.Take the exam!
EXAM DOMAINS AND WEIGHTINGS
The following table contains the domains and weightings covered on the exam. It is not a
comprehensive listing of all the content that will be presented on the exam.
Domain Domain Weightings
1.0 Snowflake for Gen AI Overview 26%
2.0 Snowflake Gen AI & LLM Functions 40%
3.0 Snowflake Gen AI Governance 22%
4.0 Snowflake Document AI 12%
Page 5
Domain 1.0: Snowflake for Gen AI Overview
1.1 Define Snowflake’s Gen AI 1.2 Outline Gen AI capabilities in
principles, features, and best practices. Snowflake.
● Snowflake Cortex ● Cortex LLM functions (for example,
○ LLMs task-specific, general)
○ Cortex Search ○ Vector-embedding
○ Cortex Analyst ○ Fine-tuning
○ Cortex Fine-tuning ● Cortex Search
○ Cortex Agents (Public ○ RAG use cases
Preview) ○ Unstructured data use cases
● Snowflake Copilot ○ REST APIs
● Security, privacy, access, and ● Cortex Analyst
control principles ○ Semantic model generation
○ Role-Based Access Control ■ Stored in YAML files in
(RBAC) a stage
○ Guardrails ■ Stored natively in
○ Required privileges semantic views (Public
○ Cortex LLM Functions Preview)
■ Control model access ○ Structured/text-to-SQL use
● CORTEX_MODELS cases
_ALLOWLIST ○ REST APIs
parameter ● Cortex Agents(Public Preview)
● Different interfaces ○ REST APIs
○ Cortex LLM Playground ● Cross-region inference
(Public Preview) ○ CORTEX_ENABLED_
○ SQL CROSS_REGION parameter
○ REST API ○ Considerations (for example,
● Different ways of bringing your latency, availability)
own models into Snowflake (for
example, from Hugging Face)
○ Using Snowflake Model
Registry (custom model)
○ Using Snowpark Container
Services
Page 6
Domain 1.0 Study Resources
Snowflake Documentation
Document AI VECTOR_L1_DISTANCE
Cortex LLM REST API VECTOR_L2_DISTANCE
Cortex Search VECTOR_COSINE_SIMILARITY
Cortex Analyst Cross-region inference
Cortex LLM Playground Fine-tuning (Snowflake Cortex)
Vector data types GRANT DATABASE ROLE
VECTOR_INNER_PRODUCT Vector Embeddings
Known limitations to Document AI
Snowflake Cortex AISQL (including LLM Reading Assets
functions) Best Practices for Getting Started with
Anomaly Detection (Snowflake ML Functions) Snowflake’s Document AI (Article)
Time-Series Forecasting (Snowflake ML Snowflake AI Trust and Safety FAQs (Article)
Functions) Snowflake’s Data Architecture: Enabling AI
Using Snowflake Copilot Apps, Next-Gen Lakehouse Analytics And
Bring your own model types via serialized files More (Article)
Snowflake Model Registry user interface
COMPLETE ([Link])
FINETUNE ('CREATE')
([Link])
Page 7
Domain 2.0: Snowflake Gen AI & LLM Functions
2.1 Apply Gen AI and LLM functions
in Snowflake. ● Choosing a model
● Snowflake Cortex ○ Considerations
○ General (e.g. capability, latency, and
■ COMPLETE cost)
■ COMPLETE
Structured 2.2 Perform data analysis given a use
Outputs case.
○ Task-specific functions ● Use fully-managed LLMs, RAG,
■ CLASSIFY_TEXT and text-to-SQL services
■ EXTRACT_ANSWER ○ Unstructured data
■ PARSE_DOCUMENT ■ CORTEX
■ SENTIMENT PARSE_DOCUMENT
■ SUMMARIZE ○ Structured data
■ TRANSLATE ○ Cortex Analyst
■ Cortex Analyst
■ EMBED_TEXT_768
Verified Query
■ EMBED_TEXT_1024
Repository (VQR)
● Cortex Search ■ Integration with
● Cortex Analyst Cortex Search
● Cortex Fine-tuning ■ Suggested Questions
● Cortex Agents (Public Preview) ■ Custom_
● Vector functions
instructions field
○ VECTOR_INNER_
PRODUCT ● Performance considerations
○ Latency (for example,
○ VECTOR_L1_DISTANCE
fine-tuning, model size)
○ VECTOR_L2_DISTANCE
○ VECTOR_COSINE_ 2.3 Build chat interfaces to interact
SIMILARITY with data in Snowflake.
● Helper functions ● Set up the Snowflake environment
○ COUNT_TOKENS ○ Required privileges
○ TRY_COMPLETE ● Invoke Cortex functions within
○ SPLIT_TEXT_ the application code (for example,
RECURSIVE_CHARACTER Streamlit)
● Chat conversations
○ Multi-turn architecture
○ Update parameters
Page 8
2.4 Use Snowflake Cortex functions in 2.5 Run third-party models in Snowflake.
data pipelines. ● Using Snowpark Container
● Snowflake Cortex Services
○ SQL interface ○ Environment setup
○ Extracting data from text ○ Docker images
using COMPLETE ○ Specification files
■ Transcripts ○ Create compute pool
○ Data enrichment ○ Create image repository
○ Data augmentation ● Using the Snowflake Model
○ Data transformations Registry
○ Logging the model
○ Calling the model
Domain 2.0 Study Resources
Snowflake Documentation
Cortex Agents TRY_COMPLETE ([Link])
Cortex PARSE_DOCUMENT CLASSIFY_TEXT ([Link])
Cortex Analyst Verified Query Repository SUMMARIZE ([Link])
Suggested Questions in Cortex Analyst EXTRACT_ANSWER ([Link])
Improve literal search to enhance Cortex [Link]
Analyst responses CREATE COMPUTE POOL
Cortex Analyst semantic model specification
Cortex Search tutorials Quickstarts and Tutorials
Dynamic tables compared to streams and tasks Fine-Tuning an LLM in Snowpark Container Services with
and to materialized views AutoTrain
AI_COMPLETE Structured Outputs Deploy and Fine-tune Open Source Llama 2 in Snowpark
Snowpark Container Services Container Services
Working with an image registry and repository Build a basic LLM chat app
Working with an image registry and repository
Snowflake Model Registry Reading Assets
Snowflake ML: End-to-End Machine Learning Snowflake Cortex Analyst: Support for Multi-turn
Model Serving in Snowpark Container Services Conversation (Article)
Snowpark Container Services Tutorials AI-Infused Pipelines with Snowflake Cortex (Blog)
SENTIMENT ([Link]) Snowpark Container Services — A Tech Primer (Blog)
TRANSLATE ([Link])
Page 9
Domain 3.0: Snowflake Gen AI Governance
3.1 Set up model access controls. 3.3 Monitor and optimize Snowflake
● Limits on which models can be Cortex costs.
used ● Cortex Search
○ Restrict access to specific ○ Different types of costs
models (virtual warehouse,
○ CORTEX_MODELS_ EMBED_TEXT, Serving)
ALLOWLIST parameter ● Cortex Analyst
■ Cortex LLM REST ○ Snowflake Service
API Consumption Table
■ COMPLETE ● Cortex LLM functions
(SNOWFLAKE. ○ Minimize tokens
CORTEX) ○ Token cost implications
■ TRY_COMPLETE ● Tracking model usage and
(SNOWFLAKE. consumption
CORTEX) ○ Usage quotas
■ Cortex LLM ○ CORTEX_FUNCTIONS_
Playground (Public USAGE_HISTORY view
Preview) ○ CORTEX_FUNCTIONS_
● Data safety and security QUERY_USAGE_HISTORY
considerations view
○ Is data leaving/going to
LLMs? 3.4 Use Snowflake AI observability tools.
● REST API authentication ● Snowflake AI observability
methods (Public Preview) features
○ Evaluation metrics
3.2 Set guardrails to filter out harmful ○ Comparisons
or unsafe LLM responses. ○ Tracing
● Cortex Guard ○ Logging
○ COMPLETE arguments ○ Event tables
● Methods to reduce model ● Implementation methods
hallucinations and bias ○ Trulens SDK
● Error conditions
Page 10
Domain 3.0 Study Resources
Snowflake Documentation
Snowflake Service Consumption Table METERING_DAILY_HISTORY view
Snowpark-optimized warehouse CORTEX_SEARCH_DAILY_USAGE_HISTORY
Evaluate AI applications view
AI Observability in Snowflake Cortex CORTEX_DOCUMENT_PROCESSING_USAGE_HI
Cortex Analyst administrator monitoring STORY view
Understanding cost for Cortex Search Services [Link]
CORTEX_FUNCTIONS_USAGE_HISTORY
view
Page 11
Domain 4.0: Snowflake Document AI
4.1 Set up Document AI. 4.3 Extract values from documents
● Virtual warehouse, database, and using Document AI.
schema considerations ● Conditions
● Roles and privileges ● <model_build_
○ USAGE name>!PREDICT query
○ OPERATE ● Automation of data pipelines
○ CREATE
[Link]. 4.4 Troubleshoot Document AI given a
DOCUMENT_ use case.
INTELLIGENCE ● Extracting query errors
○ CREATE MODEL ● GET_PRESIGNED_URL
function
4.2 Prepare documents for Document AI. ● Requirements and privileges
● Upload documents ● Cost and best practices
● Train the model considerations
● Requirements (for example,
formats, size limits)
● Question optimization best
practices
Domain 4.0 Study Resources
Snowflake Documentation Question optimization for extracting information
Document AI with Document AI
Setting up Document AI Known limitations to Document AI
Troubleshooting Document AI Extract information with Document AI
Cost governance of Document AI DOCUMENT_INTELLIGE ([Link])
Create a document processing pipeline CREATE STAGE
Prepare your documents for Document AI <model_build_name>!PREDICT
Page 12
SNOWPROⓇ SPECIALTY: GEN AI SAMPLE QUESTIONS
1. A Gen AI Specialist needs to analyze the daily costs incurred for AI services in Snowflake.
Which query will retrieve the credit consumption from Snowflake’s metadata objects for data usage?
A. SELECT * FROM SNOWFLAKE.ACCOUNT_USAGE.QUERY_HISTORY
WHERE SERVICE_TYPE='AI_SERVICES';
B. SELECT * FROM SNOWFLAKE.INFORMATION_SCHEMA.METERING_HISTORY
WHERE SERVICE_TYPE='AI_SERVICES';
C. SELECT * FROM SNOWFLAKE.ACCOUNT_USAGE.METERING_HISTORY
WHERE SERVICE_TYPE='AI_SERVICES';
D. SELECT * FROM SNOWFLAKE.ACCOUNT_USAGE.METERING_DAILY_HISTORY
WHERE SERVICE_TYPE='AI_SERVICES';*
2. What is the primary role of memory in a multi-turn chat conversation using a Gen AI model in
Snowflake Cortex Analyst?
A. To securely store user credentials
B. To increase the speed of response generation
C. To maintain context throughout multiple requests*
D. To limit the number of tokens processed for each request
3. A Gen AI Specialist is using Document AI to create a model. When creating a model build with
a name unique to the specified schema, this error is returned:
Unable to create a build on the specified database and schema.
Please check the documentation to learn more.
What would cause this error?
A. There is a model build with the same name in another schema in the database.
B. The CREATE [Link].DOCUMENT_INTELLIGENCE privilege has not
been granted to the role the Specialist is using.*
C. The USAGE privilege on the database used to create the model build has not been granted
to the role the Specialist is using.
D. The SNOWFLAKE.DOCUMENT_INTELLIGENCE_CREATOR database role has
not been granted to the role the Specialist is using.
Page 13
4. Which parameter can be used by administrators to restrict access to specific LLMs within Snowflake?
A. NETWORK_POLICY
B. SAML_IDENTITY_PROVIDER
C. CORTEX_MODELS_ALLOWLIST*
D. CORTEX_ENABLED_CROSS_REGION
5. Which Snowflake Cortex LLM function should be used to generate an instructional lesson plan
based on a prompt?
A. COMPLETE*
B. EXTRACT_ANSWER
C. SUMMARIZE
D. TRANSLATE
Page 14
MAINTAINING YOUR CERTIFICATION
All Snowflake Certifications expire two (2) years after your certification issue date.
SnowPro Certifications can now be recertified through the Snowflake Continuing Education
(CE) program which includes these options -
● Completion of eligible Snowflake Instructor Led (ILT) Training Courses
● Earning of an equivalent or higher-level SnowPro Certification
Note: You must have a valid Certification to participate in the Continuing Education (CE)
program.
The information provided in this guide is provided for your internal purposes only and may
not be provided to third parties.
IN ADDITION, THIS STUDY GUIDE IS PROVIDED “AS IS”. NEITHER SNOWFLAKE
NOR ITS SUPPLIERS MAKES ANY OTHER WARRANTIES, EXPRESS OR IMPLIED,
STATUTORY OR OTHERWISE, INCLUDING BUT NOT LIMITED TO WARRANTIES
OF MERCHANTABILITY, TITLE, FITNESS FOR A PARTICULAR PURPOSE OR
NONINFRINGEMENT.
Page 15