15 min
Lesson 1 of 4
Snowflake Cortex Fundamentals
From the course Snowflake Cortex AI
Snowflake Cortex AI brings large language models and machine learning directly into Snowflake, letting you run AI on your data without moving it to external services.
Why Cortex AI?
- Data stays in Snowflake: No data egress or external API calls for AI
- SQL-native: Call AI functions directly in SQL queries
- Governed: AI operations follow Snowflake's access controls
- Scalable: Runs on Snowflake's compute infrastructure
Core Cortex Features
LLM Functions
Call language models directly in SQL:
SELECT SNOWFLAKE.CORTEX.COMPLETE(
'mistral-large',
'Summarize this customer feedback: ' || feedback_text
) as summary
FROM customer_feedback;
Supported models: Llama, Mistral, Mixtral, Gemma, and more.
Built-in AI Functions
- SUMMARIZE(): Summarize text
- TRANSLATE(): Translate between languages
- SENTIMENT(): Analyze sentiment (-1 to 1)
- EXTRACT_ANSWER(): Extract answers from text given a question
- CLASSIFY_TEXT(): Classify text into categories
Cortex Search
Semantic search over your data:
- Index your documents and text data
- Search using natural language queries
- Returns relevant results based on meaning, not just keywords
- Powers RAG applications directly in Snowflake
Cortex Analyst
Natural language to SQL:
- Users ask questions in plain English
- Cortex generates and executes SQL queries
- Returns results as charts or tables
- No SQL knowledge required for end users
Key Takeaways
- •Cortex AI runs LLMs directly in Snowflake — data never leaves the platform
- •LLM functions are SQL-native, callable directly in SELECT statements
- •Built-in functions cover summarization, translation, sentiment, and classification
- •Cortex Search enables semantic search and RAG within your data warehouse
0 of 4 lessons marked complete
Progress is kept in this browser only.