Snowflake SnowPro Generative AI Exam Preparation

Before we explore Generative AI, we must first understand its foundations in Traditional AI. Artificial Intelligence refers to systems that can perform tasks normally requiring human intelligence, such as recognizing patterns, analyzing data, and making decisions. Traditional AI focuses on specific tasks and learns from labeled data to make predictions. While highly effective within its defined boundaries, it does not truly understand or create - it only applies learned patterns. These strengths and limitations set the stage for the next evolution of AI: Generative AI.

In this lecture, we introduce Generative AI - a new evolution of artificial intelligence that goes beyond analyzing data to actually creating new content. Unlike traditional AI, which follows fixed rules and makes predictions, Generative AI learns from vast, diverse datasets to understand context and generate novel outputs. This shift from prediction to creation allows AI to become more adaptive, creative, and collaborative, opening up entirely new possibilities in how we use intelligent systems.

In this lecture, we compare Traditional AI and Generative AI side by side to understand how they differ in purpose, behavior, and capabilities. We will see that Traditional AI focuses on analysis, classification, and prediction using deterministic rules, while Generative AI focuses on creating new, meaningful content using probabilistic methods. Through visual comparisons and real-world examples, we will explore how these two approaches handle data, adaptability, creativity, and automation, and why Generative AI represents the next major step in the evolution of AI.

This lecture provides a high-level overview of how Generative AI works, from data collection and preparation to model training, content generation, and continuous improvement through feedback. It outlines the key stages that enable Generative AI systems to learn from data, understand patterns, and produce meaningful, original outputs.

Understand Generative AI at a high level by exploring the key components across its workflow. From training data and learning to output generation and safety, this lecture highlights what powers each stage, without going into technical depth.

This lecture introduces the key Generative AI models that power different applications, including Large Language Models (LLMs), Generative Adversarial Networks (GANs), and Variational Autoencoders (VAEs).

Let's discuss why Generative AI is booming today, highlighting the role of data, advanced algorithms, and powerful computing. We also explore why Generative AI matters by illustrating its impact on productivity, innovation, customer experience, decision-making, and operational efficiency across industries.

Explore the evolution of Generative AI, from early rule-based systems like ELIZA to GANs, transformers, diffusion models, and today’s powerful multimodal AI systems.

This lecture provides a high-level overview of how Generative AI systems are structured and how their components work together. It introduces the key architectural layers, including infrastructure, data, models, prompt/orchestration, APIs, applications, governance, and monitoring. We explain the role of each layer, from supplying compute and managing data to generating outputs and delivering capabilities to end users.  By the end, learners gain a clear “big picture” understanding before diving into detailed discussions in subsequent lectures.

This lecture explains the foundational compute (infrastructure) layer that powers Generative AI systems. It highlights the role of high-performance hardware, including GPUs and TPUs, in accelerating large-scale parallel computations required for training and inference. The lecture also introduces distributed computing, showing how workloads are scaled across multiple machines to ensure performance, speed, and system

This lecture explores the data storage layer and its role in organizing, managing, and serving the data used by Generative AI systems. It explains how raw data, embeddings, and temporary results are stored efficiently to support both training and inference. The lecture also introduces feature stores, distributed storage, and checkpoints, emphasizing their importance for performance, scalability, and reliability.

At the heart of Generative AI lies the model layer, where systems learn patterns from data and generate intelligent outputs. The lecture highlights how model selection varies by task, from large language models for text to GANs for image generation. It also introduces pretraining and fine-tuning, emphasizing how existing knowledge and diverse datasets enable adaptable, high-quality results.

The model’s capabilities are shaped and directed through the prompt and orchestration layer. This lecture explains how user requests are translated into structured prompts and coordinated workflows, enabling AI systems to handle multi-step, multi-source interactions. Using intuitive analogies, it clarifies how orchestration manages sequencing, data retrieval, and response generation to deliver accurate and meaningful outputs.

Following prompt orchestration, the API layer enables applications to access Generative AI capabilities. This lecture explains how APIs act as standardized interfaces, allowing developers to integrate AI models without dealing with underlying complexity. It also introduces RESTful APIs, endpoints, and SDKs, highlighting how they simplify communication, implementation, and deployment in real-world software systems.

At the top of the Generative AI stack, the application layer converts AI capabilities into real user experiences and business value. This lecture examines how AI powers practical solutions across domains such as content creation, healthcare, and customer service. It also highlights industry-specific tools and the importance of integrating AI with business logic, workflows, scalability, and performance requirements.

With AI systems influencing real-world decisions, the governance layer provides the guardrails that ensure safe, ethical, and compliant use. This lecture explores the unique risks of Generative AI, including probabilistic outputs, bias, and potential exposure of sensitive information. It outlines how monitoring, risk controls, ethical guidelines, and auditability frameworks help maintain accountability, reliability, and regulatory alignment.

Generative AI systems require continuous oversight to remain reliable, accurate, and efficient after deployment. This lecture explains how monitoring detects performance issues, errors, and unexpected behavior, while maintenance focuses on updates, retraining, scaling, and security improvements. Together, these practices ensure AI systems stay trustworthy, resilient, and aligned with evolving data and business needs.

The next set of lectures introduces different types of Generative AI models. The lectures explain the main types of generative models and how they are used in real-world applications across business, research, and creative fields. You will also learn how these models generate text, images, audio, video, and other outputs by learning patterns from large datasets.

This lecture explains the main types of Generative AI models and how each one creates different kinds of content. You will learn about Large Language Models, GANs, VAEs, Diffusion models, and Multimodal models, and understand how they generate text, images, audio, and other data.

This lecture explains Large Language Models (LLMs), a key type of generative AI designed to understand and produce human language. You will learn how LLMs work, how they generate text using patterns learned from large datasets, and how components like instructions, context, and tools influence their responses. The lecture also covers the strengths, limitations, and real-world uses of LLMs in areas such as chatbots, content creation, automation, and software development.

This lecture explains Generative Adversarial Networks (GANs), a generative AI model that uses two neural networks working in competition to create realistic data. You will learn how the generator produces fake samples while the discriminator evaluates them, allowing both models to improve through training. The lecture also covers how GANs are used for photorealistic image generation, super-resolution, and other applications where high visual quality is important.

This lecture explains Diffusion Models, a modern generative AI approach that creates data by learning how to reverse a gradual noise process. You will learn how these models are trained by adding noise to real data and then generating new content by removing noise step by step. The lecture also covers why diffusion models produce high-quality images, how they differ from GANs, and how they are used in systems like modern AI image generators.

This lecture explains Multimodal AI models, which are designed to understand and generate multiple types of data such as text, images, audio, and video within a single system. You will learn how multimodal models combine different data sources to produce more accurate, context-aware, and human-like responses compared to single-modal models.

This lecture compares the major Generative AI model types and explains when each one should be used.

Explore the real-world challenges of building Generative AI systems, including infrastructure, scalability, operations, skills, and security. Learn how Snowflake simplifies the AI lifecycle with a unified, fully managed platform that brings data, compute, and AI together.

Learn how Snowflake simplifies Generative AI with a unified AI Data Cloud. Explore multi-interface AI development using SQL, Python, and no-code tools, while understanding how Snowflake reduces infrastructure complexity, improves scalability, and enforces built-in security and governance.

Explore the Snowflake Cortex AI Suite and learn how Snowflake unifies data, AI, analytics, applications, and governance into a single platform. Understand how enterprise data flows through AI models, enabling intelligent insights, automation, and secure AI-driven experiences.

Learn how Cortex Agents automate repetitive, rule-based tasks using AI directly inside Snowflake. Discover how agents monitor data, trigger workflows, and turn insights into real-time actions while improving efficiency, scalability, and operational control.

Discover how Snowflake Cortex AI Functions enable SQL-native AI capabilities for generating summaries, reports, images, and code directly from enterprise data. Learn how Snowflake simplifies AI adoption with fully managed, secure, and scalable AI functions.

Learn how Cortex Analyst enables users to interact with enterprise data using natural language instead of SQL. Explore how it generates insights, explains trends, and democratizes analytics by making data exploration intuitive and accessible to all users.

Explore how Cortex Fine-Tuning customizes AI models using enterprise-specific data, terminology, and workflows. Learn how organizations improve model accuracy, relevance, and performance for specialized business use cases while maintaining governance and security.

Understand how Cortex Search delivers AI-powered enterprise search across structured and unstructured data. Learn how semantic search improves relevance, understands user intent, and helps teams quickly discover insights from trusted Snowflake data.

Discover how Snowflake Copilot acts as an embedded AI assistant within Snowflake applications and workflows. Learn how it simplifies data interaction through natural language, automates repetitive tasks, and helps users work faster and smarter

Learn how Snowflake Intelligence operationalizes AI across enterprise workflows and decision-making processes. Explore how Snowflake combines governed data, analytics, and AI to deliver scalable automation, real-time insights, and intelligent business outcomes.

High-level overview of how developers build on Snowflake, introducing its layered, developer-first tools - workspaces, notebooks, Git, Snowpark, Cortex Code, Copilot, External Functions, and the Marketplace for an end-to-end AI/ML development experience.

Overview of Snowflake’s integrated IDE-style development environment - covering notebooks, workspaces, Git, Copilot, and interactive visuals, showing how developers securely build, collaborate, and iterate on GenAI and ML workflows within the platform.

Learn how Snowpark helps developers move beyond writing code by separating code authoring from execution. See how Snowflake runs data and ML workloads securely on managed compute, right next to the data for a faster, simpler workflow.

This lecture provides an overview of Snowflake Copilot and Cortex Code: Copilot lets business users ask questions in natural language and auto-generates SQL, while Cortex Code helps technical teams analyze and optimize query performance, cost, and resource usage.

This lecture provides an overview of Snowflake External Functions - how teams can securely call external AI services and models (e.g., OpenAI, AWS, Azure) from Snowflake, enabling generative AI use cases while keeping data governed, auditable, and compliant within Snowflake’s controls.

This lecture covers the Snowflake Marketplace and partnerships allowing developers to consume ready-to-use data, apps, and AI models, leverage third-party AI tools, and securely connect to cloud AI services to accelerate governed GenAI development.

This lecture outlines how Snowflake transforms enterprise data into practical AI-driven solutions. It explores key use cases, including natural language enterprise search, unified customer analytics, predictive analytics, and natural language BI. The session also highlights how AI-powered automation streamlines repetitive workflows, enabling organizations to scale AI adoption securely while maintaining governance and control.

This lecture illustrates how Snowflake’s AI capabilities adapt to different industry needs and data challenges. It explores practical examples across financial services, media and entertainment, healthcare, public sector, and marketing and sales.

This lecture provides a high-level overview of Snowflake’s machine learning capabilities, explaining how the platform supports everything from simple SQL-based ML to full end-to-end machine learning pipelines, and how these capabilities fit within Snowflake’s unified AI platform alongside Generative AI.

This lecture introduces the two primary ways to apply machine learning in Snowflake—ML Functions for quick, SQL-based predictions using prebuilt models, and ML Pipelines for building, training, and deploying scalable end-to-end machine learning workflows—helping teams choose the right approach based on their complexity and business needs.

This lecture introduces Snowflake Machine Learning Functions, explaining how teams can apply built-in, SQL-based ML capabilities directly on Snowflake data to perform forecasting, anomaly detection, classification, and insight discovery—without building, training, or managing custom machine learning models.

This lecture provides an overview of how to build, deploy, and govern custom machine learning models entirely within Snowflake, covering the full lifecycle from experimentation and automation to model management, serving, and ongoing monitoring using Snowflake ML capabilities.

In this lecture, we introduce Snowflake Cortex Code and how it uses AI to simplify SQL analysis, data exploration, documentation, and development tasks. You’ll learn its core capabilities, security model, and the two interfaces available: Snowsight for browser-based workflows and CLI for advanced local development, DevOps, and engineering use cases.

In this lecture, we explore how Snowflake Cortex Code improves data discovery and administrative workflows. You’ll learn how natural language search helps locate datasets, schemas, and documentation faster, while AI-driven administration simplifies user access, governance, security reviews, and cost visibility for more efficient Snowflake management.

In this lecture, we explore how Snowflake Cortex Code accelerates building AI agents, analytics applications, and modern data solutions. You’ll learn how it supports synthetic data generation, backend logic creation, Streamlit apps, notebook development, code migration, and optimization—helping teams build, test, and innovate faster within Snowflake.

In this hands-on lecture, you will use Cortex Code in Snowsight to generate SQL queries, debug errors, create documentation, and analyze data using natural language prompts to streamline real-world data engineering and analytics workflows in Snowflake.

In this lecture, you’ll explore Cortex Code CLI in Snowflake and learn how it accelerates development through AI-assisted workflows, automation, SQL generation, dbt integration, data diffing, and intelligent catalog search directly from the command line and modern code editors.

In this hands-on demonstration, you will install and configure the Cortex Code CLI, connect it to Snowflake using PAT authentication, and verify the setup by running commands to interact with your Snowflake environment using natural language workflows.

This lecture explains how Snowflake Cortex uses Snowflake’s Role-Based Access Control (RBAC) security model. Learn how roles manage data access, how Cortex Code inherits user permissions, how role switching works, and which mandatory roles - such as COPILOT_USER, CORTEX_USER, and CORTEX_AGENT_USER - are required to access Cortex Code features.

This lecture explains the pricing and cost management model for Snowflake Cortex Cortex Code. Learn about free trials, subscription-based CLI pricing, pay-as-you-go token billing for Snowflake customers, daily credit limits, account-level and user-level controls, and how administrators can govern Cortex Code usage across Snowsight and CLI environments.

In this hands-on lecture, you will learn how to control Cortex Code usage in Snowsight by configuring account-level and user-level credit limits, validating settings, and implementing governance controls to manage AI-related Snowflake costs.

In this hands-on demonstration, you will learn how to manage and control Cortex Code CLI credit consumption in Snowflake by configuring account-level and user-level daily usage limits. You will also verify configurations, remove limits when required, and understand how user-level overrides help enforce governance and cost control in real-world enterprise environments

In this lecture, we compare Snowflake’s three generations of AI tools: Cortex Code, Snowflake Intelligence, and the legacy Snowflake Copilot. You’ll learn how their purpose, task coverage, capabilities, and design philosophy differ, and how Snowflake has evolved from basic SQL assistance to advanced AI agents supporting engineering, analytics, and operations.

Explore the key features of Cortex Playground, including prompt experimentation, model parameter tuning, side-by-side model comparison, and integration with Snowflake tables for testing against real enterprise data. Also learn about the required CORTEX_USER role and how Snowflake governs secure access to AI capabilities.

Compare and optimize AI-generated responses using Snowflake Cortex Playground. In this hands-on lab, you will evaluate multiple language models, experiment with parameters like temperature and max_tokens, and analyze how different configurations impact response creativity, clarity, reasoning depth, and completeness for real-world enterprise analytics use cases.

Discover how Snowflake CoWork (previously known as Snowflake Intelligence) enables users to interact with enterprise data using natural language. In this lecture, you will explore how business users can ask conversational questions, analyze trends, uncover business insights, and combine internal and external data to support faster, AI-powered decision-making across the organization.

Understand how Snowflake CoWork (previously known as Snowflake Intelligence) processes natural language questions using Cortex Agents, orchestration, and AI-powered tools to deliver governed business insights. In this lecture, you will explore how Snowflake CoWork converts user questions into actionable responses using semantic models, Cortex Search, Cortex Analyst, visualizations, and secure enterprise data access.

iscover how to use Snowflake CoWork to analyze enterprise data with natural language. Learn how to interact with an existing Cortex Agent, ask business questions, generate SQL behind the scenes, and create visualizations without writing SQL.

Explore Cortex Analyst, Snowflake’s fully managed text-to-SQL solution. Learn how natural language questions are transformed into SQL, executed against Snowflake data, and returned as business insights. Discover semantic models, LLMs, REST APIs, and how Cortex Analyst simplifies conversational analytics at scale.

Explore the end-to-end architecture of Cortex Analyst and learn how natural language questions are transformed into SQL. Follow the complete request flow, including question understanding, semantic context enrichment, SQL generation, query execution, and governed access through Snowflake roles, Semantic Views, and Semantic Models.

Discover the key capabilities of Cortex Analyst, including natural language analytics, REST API integration, advanced LLM-powered SQL generation, semantic-layer-driven accuracy, privacy-first architecture, and enterprise governance. Learn how these features enable secure, scalable, and accurate conversational analytics experiences.

Semantic Views provide the business context that powers Cortex Analyst. Explore logical tables, dimensions, facts, metrics, and relationships, and discover how rich metadata, business rules, predefined joins, and verified query examples help improve SQL generation accuracy and conversational analytics experiences.

Learn how Cortex Analyst supports multi-turn conversations by maintaining context across multiple questions. Discover how follow-up queries are interpreted using previous interactions, enabling more natural, efficient, and conversational data exploration without repeatedly restating business context.

Learn the access control requirements for Cortex Analyst. Explore the roles, privileges, and permissions needed to use Cortex Analyst, access semantic models, Cortex Search Services, and underlying tables, while ensuring all AI-generated queries comply with Snowflake's security and governance framework.

Learn how to build a Cortex Analyst using Snowsight. Configure a semantic view with Autopilot, provide SQL context, review tables and columns, and test natural language queries in the Playground to deliver self-service analytics without writing SQL.

Discover what AI agents are and how Cortex Agents extend Snowflake's AI capabilities beyond individual services. Learn how agents understand goals, plan tasks, orchestrate multiple tools, maintain context, and coordinate actions across structured and unstructured data to solve complex business problems.

Explore the key challenges of building enterprise AI agents, including siloed data, fragmented technology stacks, governance, complex workflows, and operational costs. Learn why these obstacles require a unified orchestration approach and how Cortex Agents help address them.

Explore the architecture of Cortex Agents and learn how the orchestration layer coordinates language models, Cortex Analyst, Cortex Search, and custom tools. Discover how Agent Objects, enterprise security, and conversation threads enable governed, context-aware AI across structured and unstructured data.

Learn the access control model for Cortex Agents and the roles required to use them securely. Explore the differences between SNOWFLAKE.CORTEX_USER and SNOWFLAKE.CORTEX_AGENT_USER, and discover how role-based access control supports secure, governed AI access using the principle of least privilege.

Build your first Cortex Agent in Snowsight by configuring an Agent Object, adding instructions and sample questions, integrating Cortex Analyst and Cortex Search tools, and testing how the agent answers natural language questions using structured and unstructured data.

In this series of lectures, we’ll explore Snowflake’s powerful text analysis and summarization capabilities through a set of built-in AI functions: SUMMARIZE, AI_SENTIMENT, AI_EXTRACT, and AI_CLASSIFY. These functions make it easier to work with unstructured data by turning raw text into meaningful insights. We’ll begin with a clear overview of what each function does and where it can be applied. Then, we’ll move into hands-on demonstrations using real datasets, where you’ll learn how to generate concise summaries, detect sentiment, extract structured details, and classify text efficiently within Snowflake.

Learn how Snowflake’s SUMMARIZE function converts long text into concise summaries. Explore syntax, real-world use cases, and a practical example using customer feedback analysis.

In this hands-on demo, learn how Snowflake’s SUMMARIZE function helps convert long customer feedback into concise summaries, enabling faster issue identification, prioritization, and improved support efficiency.

Learn how Snowflake’s AI_SENTIMENT function detects positive, negative, neutral, or mixed sentiment. Explore overall and aspect-based analysis with practical customer review examples.

Learn how to use Snowflake Cortex AI_SENTIMENT() to evaluate the emotional tone of text. Analyze customer feedback as positive, negative, or mixed and understand JSON output structure.

Learn how Snowflake’s AI_EXTRACT function converts unstructured text into structured data. Explore syntax, use cases, and examples for extracting key fields, lists, and insights.

Learn how to use Snowflake Cortex AI_EXTRACT() to transform unstructured customer feedback into structured insights. In this hands-on activity, extract key details such as issue type, product, urgency, and customer intent using practical real-world examples.

Learn how Snowflake’s AI_CLASSIFY function categorizes text and images using custom labels. Explore syntax, single-label and multi-label classification, configuration options, and practical business examples.

Learn how to use Snowflake Cortex AI_CLASSIFY() to automatically classify text into predefined categories. In this hands-on activity, apply custom labels to customer feedback, understand classification outputs, and explore how AI-driven categorization improves automation and decision-making.

Generative AI brings a distinct set of enterprise security and risk management challenges that extend beyond traditional data protection. This lecture walks through the end-to-end AI/ML lifecycle, highlighting where risks can emerge, ranging from source data and enrichment to model development, deployment, and retirement. Learners gain a structured understanding of threat vectors such as data leakage, model poisoning, jailbreaks, drift, and governance gaps, reinforcing the need for layered security and continuous oversight.

Network security serves as the first line of defense in protecting Generative AI workloads within Snowflake. This lecture explains Snowflake’s zero-trust, locked-down connectivity model, including network policies, private endpoints, and controlled inbound access. It also highlights internal workload isolation and governed outbound connections, showing how these mechanisms reduce risks such as unauthorized access, data leakage, and exfiltration.

Identity and Access Management in Snowflake controls who and what can access data, models, and AI services. This lecture covers authentication mechanisms such as SSO, MFA, OAuth, and key-based access, along with integration with enterprise identity providers to enforce consistent security policies. It also explains how RBAC, least-privilege roles, and model allowlisting protect against risks like credential misuse, model poisoning, exfiltration, and uncontrolled compute consumption.

Understanding data is a prerequisite for governing and securing Generative AI systems. This lecture explains how classification, tagging, and lineage provide visibility into data sensitivity, usage, and movement across AI workloads. It also shows how these capabilities support consistent policy enforcement, risk mitigation, compliance, and the prevention of issues such as data leakage, poisoning, and broken pipelines.

Snowflake’s data protection framework secures sensitive information through layered controls such as RBAC, column-level security, dynamic masking, and tokenization. This lecture highlights how tagging, row access policies, views, and UDFs minimize unnecessary data exposure while maintaining usability. It also connects these protections to Generative AI risk mitigation, addressing threats like data exfiltration, model misuse, shadow ML, and compromised data pipelines.

This lecture introduces the Model Registry as a critical control point for securing machine learning and Generative AI workflows. It explains how a governed registry mitigates risks such as data poisoning, model exfiltration, unauthorized access, and use of obsolete models through encryption, RBAC, auditing, and monitoring. Learners see how a structured model lifecycle strengthens security, compliance, and operational trust across AI systems.

This lecture explains how Snowflake unifies governance and security controls across both training and inference stages of the Generative AI lifecycle. It highlights protections such as strict access control, encryption, lineage tracking, and Cortex Guard to ensure models are trained on trusted, validated data. The session also covers inference-time safeguards, including controlled prompt access and AI Observability, to monitor model behavior, prevent drift, and maintain compliant, reliable outputs.

The next set of lectures introduces how access to AI capabilities is governed in Snowflake Cortex. You will learn the difference between coarse-grained and fine-grained access control, and how allow lists and RBAC work independently and together in a hybrid model. We also cover governance considerations for AI models, cross-region access, and securing access to the Cortex REST API. By the end, you will understand how to design controlled, secure, and practical access policies for enterprise AI.

This lecture explains the two primary access control mechanisms: coarse-grained and fine-grained access control. You will learn how coarse-grained access provides broad permission buckets, while fine-grained access controls permissions at a detailed level. Through a practical office building analogy, you will understand the benefits, limitations, and when to use each approach. By the end, you will clearly understand how these mechanisms support secure and flexible governance.

This lecture introduces the two mechanisms used to control AI model access in Snowflake Cortex: the account-level allow list and role-based access control (RBAC). You will learn how the allow list acts as a coarse-grained control that defines which models are available across the account, while RBAC provides fine-grained control over who can use those models. We then explore the CORTEX_MODELS_ALLOWLIST parameter, its configuration options, and how it establishes governance boundaries before detailed permissions are applied.

This lecture introduces role-based access control (RBAC) as the primary mechanism for controlling who can access and use AI models in Snowflake Cortex. You will learn how privileges are assigned to roles, how users inherit permissions, and how RBAC enables fine-grained control over model usage. We then examine how RBAC works together with the CORTEX_MODELS_ALLOWLIST parameter, including scenarios where both controls are applied.

In this hands-on lab, you will configure the account-level Cortex models allowlist parameter to control which AI models are permitted for use. You will add models to the allowlist, refresh the model metadata, and test model execution using the AI COMPLETE function. You will first validate that you can use allowed models, and then attempt to call a model that is not included in the allowlist. This will help you observe how Snowflake blocks unapproved models and enforces governance. By the end of this lab, you will understand how to manage model access using the allowlist parameter.

In this hands-on lab, you will disable AI model access by configuring the CORTEX_MODELS_ALLOWLIST parameter to None. You will modify the account-level setting, execute an AI model query, and observe how Snowflake blocks model usage when no models are approved. This lab demonstrates how administrators can restrict AI model access across the account and enforce governance controls.

This lecture explains how Snowflake prepares Cortex models for RBAC-based access control. You will learn how running the CORTEX_BASE_MODELS_REFRESH procedure synchronizes available models and creates securable objects in the SNOWFLAKE.MODELS schema. We also cover how Snowflake generates application roles for each model, how these roles are used to control access, and how users reference models in queries. This lecture focuses on concepts and architecture to help you understand how RBAC enforcement works.

Configure RBAC to control access to Cortex AI models when the account allowlist is disabled. Create model objects, grant application roles, assign user roles, and verify that only authorized users can access specific foundation models.

In this hands-on lab, you'll explore the different methods for referencing Cortex AI models. You'll list available model objects, invoke models using fully qualified and partial identifiers, and use automatic model resolution to simplify AI function calls.

Learn how Snowflake secures Cortex REST API access through authentication and authorization, follow the complete end-to-end lifecycle of a REST API request. We’ll also request processing, prompt caching, model inference and responses.

In this lecture, you will learn how bias emerges in Generative AI systems, why it creates ethical and business risks, and how Snowflake Cortex Guard helps enforce safer, fairer, and more responsible AI-generated outputs.

This lecture demonstrates how bias can emerge in Generative AI from both biased and neutral prompts, showing how training data, model responses, and feedback loops can reinforce stereotypes and unfair outcomes over time.

This lecture explores the real-world consequences of bias in Generative AI, including unfair decisions, reinforced stereotypes, and loss of trust, highlighting why responsible AI governance is critical for organizations.

This lecture explains hallucinations in Generative AI, how models generate confident but incorrect responses, and why hallucinations create serious risks for trust, compliance, reliability, and decision-making in enterprise systems.

This lecture explains why Generative AI hallucinations occur, including limited training data, memorization, ambiguous prompts, and language interpretation issues that cause models to generate confident but incorrect responses.

This lecture explores content safety challenges in Generative AI, including harmful outputs, real-time moderation, scalability, cultural context, and the limitations of AI guardrails when responding to risky, unethical, or potentially harmful prompts.

This lecture introduces Generative AI guardrails and explains how safety controls, validation checks, and filtering mechanisms help prevent hallucinations, bias, and harmful outputs while ensuring AI systems remain safe and reliable.

This lecture explains how bias can enter Generative AI systems across the AI lifecycle—from data collection and model design to optimization and deployment—and why continuous governance is essential for responsible AI.

This lecture demonstrates how AI guardrails improve safety by preventing Generative AI systems from providing harmful or inappropriate responses, ensuring outputs remain responsible, controlled, and aligned with real-world safety requirements.

This lecture introduces Snowflake Cortex Guard, the built-in Generative AI guardrails framework that helps reduce harmful outputs, enforce enterprise governance policies, and moderate AI responses using automated real-time safety checks.

This lecture explores the benefits of Snowflake Cortex Guard, including scalable AI guardrails, protection against unsafe prompts, reduced operational complexity, low-latency moderation, and enforcement of enterprise compliance and governance policies.

Learn how Snowflake AI COMPLETE brings generative AI directly into SQL. In this lecture, you'll explore how to create intelligent text responses, leverage enterprise data in prompts, design effective instructions, produce structured outputs, and select the right AI model for your use case.

Discover how to control and secure AI COMPLETE responses using parameters such as Cortex Guard. In this lecture, you'll learn how guardrails filter unsafe content, protect sensitive information, and improve AI safety through a real-world customer support scenario and hands-on lab demonstration.

Learn how Cortex Guard strengthens AI safety within the AI_COMPLETE function. In this hands-on lab, you'll compare responses with and without Guardrails enabled, exploring scenarios involving sensitive personal information and misinformation while demonstrating how Cortex Guard helps enforce safer, compliant, and responsible AI outputs.

The Temperature parameter controls the creativity and randomness of AI-generated responses. Lower values produce focused, consistent, and reliable outputs, while higher values encourage more diverse and creative responses. In Snowflake Cortex AI_COMPLETE function, temperature helps balance precision and exploration, though it may also increase the risk of hallucinations.

Learn how the Temperature parameter influences AI-generated responses in Snowflake AI_COMPLETE. Compare outputs generated at different temperature settings and observe the trade-off between consistency, predictability, creativity, and response variability in customer support scenarios.

Learn how the Top-P parameter influences AI_COMPLETE responses by limiting the pool of likely words the model can choose from. Through a practical customer support example, see how lower and higher Top-P values balance focused, reliable outputs with greater flexibility and creativity.

Explore how the Top-P parameter influences AI-generated content in Snowflake AI_COMPLETE. Compare outputs generated with different Top-P values and observe how token selection impacts response consistency, creativity, diversity, and overall content style.

Compare Temperature and Top-P to understand when each should be used in AI_COMPLETE. Learn how these parameters influence creativity, consistency, and word selection, and discover practical guidelines for choosing the right combination for analytical and creative AI tasks.

Learn how the Max Tokens parameter controls the length of AI_COMPLETE responses. Through a practical customer support example, see how values that are too low can truncate responses, while higher values can produce unnecessary detail, helping you choose the right balance.

Experiment with the Max Tokens parameter in Snowflake AI_COMPLETE by generating marketing content using different token limits. Compare outputs at 100, 200, and 500 tokens to observe truncation, completeness, and identify the optimal response length for business scenarios.

Learn how the response_format parameter controls the structure of AI_COMPLETE outputs. Explore how JSON schemas can enforce consistent, machine-readable responses for automation, while also generating business-friendly formats for human consumption, making AI outputs reliable for both applications and users.

Learn how to use Structured Responses in AI_COMPLETE to generate predictable JSON output. Compare free-form and structured responses, define a custom JSON schema, and produce machine-readable results that integrate seamlessly with applications and automation workflo

Learn how the AI_SQL_ERROR_HANDLING_USE_FAIL_ON_ERROR session parameter improves the resilience of AI_COMPLETE workloads. Discover how disabling fail-on-error allows batch processing to continue despite individual record failures, returning structured error details while successfully processing all valid records.

This lecture introduces Snowflake AI observability concepts, including applications, external agents, versions, datasets, runs, metrics, and traces, helping you monitor, evaluate, and debug AI systems at scale.

This lecture provides a deeper exploration of Snowflake AI observability concepts, explaining how Applications, External Agents, and Versions help organizations monitor, evaluate, govern, and compare Generative AI workflows across evolving models, prompts, orchestration logic, and production environments.

This section explains the core observability concepts used in Snowflake AI Observability, including Datasets, Runs, Metrics, and Traces, and how these components work together to monitor, evaluate, and improve AI application performance.

Learn how Snowflake AI Observability uses TruLens to monitor and evaluate LLM applications. Explore application instrumentation, telemetry capture, evaluation datasets, governed metric computation, Snowsight dashboards, and side-by-side comparison of AI application runs for continuous optimization.