Beyond Chatbots: In-Database GenAI for Advanced Analytics with SQL

This session moves beyond familiar LLM applications to demonstrate a powerful, in-database approach for advanced analytics. We’ll dive into BigQuery ML’s AI.GENERATE_TABLE function, showcasing how you can run Gemini models directly on your data warehouse using simple SQL. We’ll walk through end-to-end notebooks that tackle real-world business problems: extracting structured features like sentiment and urgency from unstructured patient surveys, performing unsupervised theme discovery on customer support chats, and generating rich customer personas from raw transaction data.

The core of the demo will highlight how these newly generated features can be immediately joined with existing structured data (e.g., clinical records, sales figures) to power deeper analysis and predictive models—all within a unified, zero-ETL workflow. We will also briefly showcase how Colab’s built-in Data Science Agent can be used to accelerate the development and analysis of these powerful notebooks.