Snowflake Cortex AI: Applied AI with SQL and LLMs

Snowpark ML – The Engineer’s Toolbox

• Snowpark architecture and execution model vs. traditional ML workflows.
• DataFrames and distributed computing in Snowflake.
• Feature engineering pipelines and model training using Python inside Snowflake.
• Decision Framework: Choosing between Snowpark ML (Custom) and Cortex AI (Built-in).
• Integration with external libraries (Sklearn, XGBoost) and architectural constraints.

Getting Started with Cortex AI

• Introduction to the SQL-first AI paradigm.
• Use-case-driven AI development and Cortex classification wizards.
• Understanding ML Classes, Methods, and the object lifecycle.

ML-Powered Functions and SQL Classes

• SQL Class Instances and the structure of Cortex ML functions.
• Invoking ML functions via SQL and managing model objects.
• Comparison: ML-as-a-Service vs. Cortex native built-ins.
• Use Case: Step-by-step SQL-based ML workflow implementation.

Classification Models in Action

• Binary and Multi-class classification.
• Business Scenario: Loan approval prediction.
• Evaluating outputs: Confusion matrices and heatmap interpretation.

Time Series Forecasting with Cortex

• Forecasting fundamentals and data preparation for time-series.
• Training models for sales and temperature prediction scenarios.
• Differences between traditional ML forecasting and Cortex SQL-based forecasting.

Anomaly Detection Made Simple

• Detecting outliers in sales and IoT/temperature data.
• Automating anomaly alerts using Snowflake Tasks and Alerts integration.

Gradient Boosting in Cortex

• Gradient Boosting fundamentals and classification vs. regression boosting.
• Performance comparison and business case examples.

Contribution Explorer – Explainability in Action

• Feature contribution analysis and explaining prediction shifts.
• Sales driver analysis and Titanic survival visual explanations.

Managing Access and Cost

• Roles and permissions for ML classes and secure model execution.
• Governance controls, cost estimation strategies, and budget planning for AI.

LLM Functions in Cortex

• Generative AI fundamentals and LLM vs. Traditional ML decision-making.
• Core Functions: SENTIMENT, COMPLETE, EXTRACT_ANSWER, SUMMARIZE, and TRANSLATE.
• Use Cases: Customer feedback analysis and multi-language document processing.

ChatGPT and Snowflake Integration

• Local ChatGPT applications vs. Snowflake-integrated LLM workflows.
• Secure LLM invocation patterns, access control, and cost management.

Extensions, Copilot, and Document AI

• Snowflake Copilot demonstration and Universal Search.
• SQL generation using LangChain and ChatGPT; reliability and validation.
• Document AI architecture and private document intelligence using LlamaIndex.