Snowflake is set to offer customers a managed service that securely unleashes AI models on enterprise data. With Cortex, Snowflake says it is giving everyone the opportunity to generate value from generative AI, regardless of expertise. The company itself already has a few “LLM-powered experiences” ready to go in Private Preview: Document AI, Snowflake Copilot and Universal Search.
Snowflake Cortex’s promise is all about simplicity. SVP of AI at Snowflake Sridhar Ramaswamy emphasizes that customers can build powerful and secure applications around their data. “With Snowflake Cortex, businesses can now tap into the power of large language models in seconds, build custom LLM-powered apps within minutes, and maintain flexibility and control over their data — while reimagining how all users tap into generative AI to deliver business value.”
Develop
Snowflake says it gives companies all the “building blocks” needed to deploy LLMs and AI without having to set up expensive GPU-based infrastructure. Within Cortex, all kinds of data can be processed and analyzed in the most efficient way possible. Two serverless functions available via a function call in SQL or Python are available in Private Preview. Specialized Functions can discover the sentiment behind an input, provide a response, summarize text and translate into a chosen language. The functionality of existing and soon-to-be available ML capabilities within Snowflake is also supported through Specialized Functions, such as anomaly detection and classification.
Snowflake also uses well-known open-source LLMs, such as Llama 2 from Meta. A text-to-SQL language is also available to “talk” to data for a variety of applications, as Snowflake characterizes it.
Streamlit in Snowflake has been running in Public Preview for some time, where LLM-driven apps can be built in Python. More than 10,000 apps are said to have already been developed with it, with applications in the medical, customer service and credit reporting industries.
Snowflake’s new apps
The aforementioned “LLM-powered experiences” would be ideal for business teams and analysts. Snowflake Copilot is a new AI assistant (in Private Preview) for writing SQL queries and gaining insights into data in plain language. There are now numerous Copilots to name, but the Splunk AI Assistant announced in mid-2023 is most reminiscent of it.
Universal Search is also launching in Private Preview and should allow customers to see relevant data and applications faster. One can search across an organizations Snowflake account, including databases, views and so-called Iceberg Tables.
With Document AI, organizations can more quickly search invoices, contracts and other documents for relevant information. In fact, it allows information that is difficult to track through classifications to become instantly insightful. Again, it is possible to refine results with dead words; again, this is a Private Preview for now. Snowflake describes Document AI as the company’s “first LLM experience.”
Customizability
Finally, Snowflake cites Snowpark Container Services, a service that simplifies management of containerized workloads. This tool also allows external LLMs to be securely run in a separate container within Snowflake’s own environment. Any open-source LLM can also be customized as desired within this service.
Also read: Snowflake acquires Ponder for better integration Python workflows