Hands-On Workshop: Building AI Functions with Starburst Galaxy
As enterprises rush to operationalize AI, most teams encounter the same challenge: connecting large language models to real business data without introducing layers of complexity, security risks, or new infrastructure. Data and AI practitioners are forced to move data out of governed environments just to experiment with prompts or retrieval workflows—slowing innovation and breaking compliance in the process.
In this hands-on workshop, you’ll see how you can eliminate those barriers using AI Functions in Galaxy. Together, we’ll configure a model, utilize Starburst functions to prompt it from SQL, apply task-based AI functions (such as sentiment analysis, classification, translation, and masking), and generate embeddings for vector search. Finally, you’ll build a simple RAG workflow, using only SQL CTEs and similarity functions, to retrieve relevant context and enrich prompts directly inside the Galaxy query editor.
You’ll walk away with practical techniques for:
- Configuring models and privileges for AI workflows
- Invoking LLMs from SQL via prompt and task functions
- Creating and storing vector embeddings in Iceberg tables
- Building a minimal RAG pipeline entirely in SQL
Register
As enterprises rush to operationalize AI, most teams encounter the same challenge: connecting large language models to real business data without introducing layers of complexity, security risks, or new infrastructure. Data and AI practitioners are forced to move data out of governed environments just to experiment with prompts or retrieval workflows—slowing innovation and breaking compliance in the process.
In this hands-on workshop, you’ll see how you can eliminate those barriers using AI Functions in Galaxy. Together, we’ll configure a model, utilize Starburst functions to prompt it from SQL, apply task-based AI functions (such as sentiment analysis, classification, translation, and masking), and generate embeddings for vector search. Finally, you’ll build a simple RAG workflow, using only SQL CTEs and similarity functions, to retrieve relevant context and enrich prompts directly inside the Galaxy query editor.
You’ll walk away with practical techniques for:
- Configuring models and privileges for AI workflows
- Invoking LLMs from SQL via prompt and task functions
- Creating and storing vector embeddings in Iceberg tables
- Building a minimal RAG pipeline entirely in SQL
