Bozeman, Montana-based DaaS firm Snowflake has partnered with Nvidia to let clients customize LLMs (large language models) using proprietary data in the Snowflake Data Cloud. Nvidia’s NeMo platform and GPU-accelerated computing will power the effort to tailor models to specific business use cases, such as chatbots with category expertise as opposed to generalists, search engines attuned to context or generative text deep knowledge. Since most companies are eager to harness brand-specific AI without having to build a model from scratch, this category of service is generating a lot of interest.

“Together, Nvidia and Snowflake will create an AI factory that helps enterprises turn their valuable data into custom generative AI models to power groundbreaking new applications — right from the cloud platform that they use to run their businesses,” Nvdia founder and CEO Jensen Huang said in an announcement issued during Snowflake Summit 2023.

Manuvir Das, head of enterprise computing at Nvidia, used an analogy that likened AI models to employees, with a foundation model equivalent to a recent college graduate, while a customized model has the chops of a 20-year vet. “The long-time employee has acquired the institutional knowledge needed to solve problems quickly and with accurate insights,” Das told VentureBeat.

The joint initiative effectively sees Nvidia embedding NeMo in the Snowflake Data Cloud. NeMo is a cloud-native platform for building, customizing and deploying generative AI models with billions of parameters. Proprietary data “can range from hundreds of terabytes to petabytes of raw and curated business information,” according to Nvidia.

Snowflake also announced is it working with partner Microsoft to make Azure OpenAI records easier to use when stored on the Snowflake Data Cloud. The Azure OpenAI Service provides OpenAI language models that can be customized.

SiliconANGLE cites ways clients might want to cross-pollinate data between the two firms, writing that “a company that uses Snowflake to store sales logs, for example, could leverage a language model from Azure OpenAI Service to automatically determine which of its products are most popular.” Or data stored on Snowflake could be used for neural network training.

Microsoft’s May Azure OpenAI update, Azure AI Studio, lets companies train OpenAI models on custom datasets. Now in preview, it’s garnered some feedback for ease of use.

Related:
Reka AI Raises $58 Million to Customize LLMs for Enterprise, ETCentric, 6/30/23
OneTrust Integrates with Snowflake to Automate Data Discovery, Classification, PR Newswire, 6/28/23
Kumo Empowers Deep Learning in Snowflake Data Cloud, VentureBeat, 6/27/23
RelationalAI and Snowflake to Revolutionize Enterprise AI Decision-Making, VentureBeat, 6/27/23
Snowflake Launches LLM-Driven Document AI and More at Annual Conference, VentureBeat, 6/27/23
Snowflake’s Finding NeMo to Train Custom AI Models, The Register, 6/27/23