BigQuery Meets LLM: Unlocking New Frontiers in AI-Driven Data Analytics

Unlocking a level of understanding that was previously unimaginable.

Photo by Shoeib Abolhassani on Unsplash

Introduction

This started when I wanted to create embedding vectors for certain categorical items for machine learning features. Upon exploring many possible options I found that BigQuery already supports embedding vectors (still in preview when I write this story).

What is BigQuery? BigQuery, in simple terms, is a fully-managed, serverless data warehouse provided by Google Cloud. It is designed to analyze and query large datasets using SQL (Structured Query Language)

What are LLMs? LLMs, or Large Language Models, are AI wizards trained on vast amounts of text and code. Their expertise lies in understanding natural language, generating human-quality text, and performing various tasks like translation, summarization, and code completion.

So why not explore it for more details? BigQuery uses LLM in Vertex AI using an external connection and accesses it through API. Once you have created a remote model over the AI resource, you access that resource’s functionality by running a BigQuery ML function against the remote model.