Snowflake Cortex AI: A deep dive into its features
Snowflake launched Cortex AI (launch date not specified in the provided source), a tool that helps businesses use powerful language models, for many inference tasks without requiring customers to provision GPUs or other infrastructure; however, some advanced training and serving features support GPU-powered containers.
Everything runs inside Snowflake with simple SQL or Python, so you can generate text, analyze sentiment, translate languages, summarize info, and more—all in one place.
Ask questions about your data in plain English
Cortex AI comes packed with features: you can ask questions about your data in plain English (even across different types of files), pull details from things like invoices using Document AI, and search documents smarter with tools like retrieval-augmented generation (RAG).
Fine-tune language models for your needs
You can fine-tune these language models for your own needs; Snowflake provides serverless fine-tuning for many workflows, while larger training or specialized serving can use managed GPU/container compute.
New inference model options include Llama 3.1 and 3.2; fine-tuning is supported for select base models and may not include every available inference model.
Plus, Cortex Guard helps block harmful content while AI Observability lets you track how well everything's working behind the scenes.