Preparing for the Snowflake Gen AI Certification: Breaking Down the GES-C02 Study Guide

If you are planning to take the SnowPro® Specialty: Gen AI (GES-C02) certification, one of the first things I would recommend is:

Don’t just start studying the topics. Start by understanding the exam.

I’m preparing for the certification myself, so I decided to go through the official study guide, break down the domains, and create a focused study plan.

The previous SnowPro® Specialty: Gen AI (GES-C01) exam has been retired, and the current exam is SnowPro® Specialty: Gen AI (GES-C02). The domain weightage is changed, and highest weightage is now on Gen AI Functions, followed by Gen AI Governance. So, instead of giving equal time to every topic, I’m planning my preparation based on the exam weightage.

First, let’s look at the exam

Total Questions: 55
Question Types: Multiple Choice & Multiple Select
Time Limit: 85 minutes

Based on the published domain weightings, the approximate number of questions would be:

Domain weightage based on Study Guide

The actual distribution of questions may vary, but this gives us a good idea of where to spend our preparation time.

And this immediately tells us something:

Gen AI Functions + Gen AI Governance = 67% of the exam

That’s where I want to spend most of my study time — especially with hands-on practice.

So, let’s break down each domain.

Get Minal Bhatkar Satpute’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

Domain 1 — Gen AI Overview

• This domain has 18% weightage, so we can expect roughly 10 questions from this area.
• This section is mainly about understanding the overall Gen AI capabilities available in Snowflake. They will judge how well we understand the different Gen AI features and, more importantly, where they fit.
• Since Cortex Code is relatively new and now well established, I think we should pay special attention to it and understand the different ways we can use it.
• Another important area is Snowflake Cortex — including Cortex Models and Functions, Cortex Search, Cortex Analyst and Cortex Agents. We should understand what each of these does and the kind of use cases they are designed for.
• We should also cover Snowflake Intelligence and the different ways we can interact with Snowflake’s AI capabilities, such as AI Studio, SQL and REST APIs.
• BYOM is another area to understand, especially Snowflake Model Registry and Snowpark Container Services.
• And finally, I would spend some extra time on Cortex Analyst, particularly Semantic Views, as this is an important part of the capabilities covered in this domain.

Domain 2 — Gen AI Functions

• This is the highest-weighted domain at 38%, so we can expect roughly 21 questions from this section. Clearly, this is where we need to spend the most time.
• The first thing I want to focus on is the different Snowflake Cortex AI functions — general functions, task-specific functions, helper functions and vector functions. But I don’t want to just memorize the function names or syntax. The important part is to understand which function should be used for which scenario. Given a particular use case, we should be able to identify the right function and understand why it is the right choice.
• The domain also covers how we can use Gen AI capabilities for data analysis, including functions for sentiment analysis, summarization, translation, embeddings, filtering, similarity, transcription and redaction.
• Then we move into some more practical areas — using fully managed LLMs, RAG and text-to-SQL, working with structured and unstructured data, and understanding how Cortex Search and Cortex Analyst fit into these use cases.
• We should also understand how to build applications that allow users to chat with data in Snowflake, including multi-turn conversations and using Cortex functions from application code.
• And finally, we need to know how to run third-party models in Snowflake using Snowpark Container Services and the Snowflake Model Registry. This is definitely going to be one of the areas where I want to do hands-on practice rather than just reading the documentation.

Domain 3 — Gen AI Governance

• This domain has 29% weightage, which means roughly 16 questions. Interestingly, this domain has gained more weight compared to the previous version of the exam, so I think we need to take this section quite seriously.
• This section is essentially about governance, security, access control, cost management and observability around Gen AI workloads.
• First, we need to understand how we control which models can be used and who can access them. This includes application roles, RBAC, account-level allowlists and the different privileges required for Cortex services.
• Security is another major area. We need to understand things like cross-region inference, guardrails, sensitive data management, AI_REDACT, authentication methods, and approaches that can help reduce hallucinations and bias.
• Then comes cost management. We need to understand what contributes to Cortex costs, how token usage impacts cost, how to monitor model consumption and which usage history views can help us track it.
• And finally, we have AI observability — evaluation metrics, comparisons, tracing, logging, event tables and implementation using tools such as the TruLens SDK.
• Given the weightage of this domain, I definitely don’t want to treat governance as something we study at the end. It needs to be part of the preparation from the beginning.

Domain 4 — Document Processing

• This is the final domain, with 15% weightage, so we can expect roughly 8 questions. This domain is focused on how we can process and extract information from documents using Snowflake’s Gen AI capabilities.
• We first need to understand the document parsing functions, particularly AI_PARSE_DOCUMENT and AI_EXTRACT, and what each of them is used for.
• Then we move into preparing and managing documents — understanding supported formats, size limitations and other requirements before we start processing them.
• The interesting part for me is how we can turn this into an automated workflow. We need to understand how Cortex AI functions can work together with Snowflake features such as Streams and Tasks to build document processing pipelines.
• And finally, we need to look at troubleshooting and optimization — things like query errors, required privileges, token usage, document batching, cost considerations and fine-tuning extraction models.
• This may have the lowest weightage of the four domains, but there are still around 8 questions here, so I definitely wouldn’t skip it.

So, what’s my preparation strategy?

For my preparation, I don’t want to simply go through the study guide and memorize the features.

Whenever I pick up a new topic, I want to first understand what it is and what problem it solves. Then I want to see where I would actually use it in a real-world scenario. And wherever possible, I want to get my hands dirty and actually try it myself.

Once I feel comfortable with the topic, I’ll come back to it during revision and make sure I can still connect the concept with the right use case.

And that’s exactly how I’m planning the next posts in this series.

I’ll be breaking down each domain into smaller topics, sharing my notes, practical examples, hands-on exercises, and resources that I find useful while preparing.

So if you’re also preparing for GES-C02, follow along.

Let’s learn Gen AI together — one domain at a time. 🚀

And if you’ve already taken the exam or are preparing for it, I’d love to know:

What topic do you think deserves more attention than it gets in the study guide?