Snowflake and Nvidia partnership could make building generative AI applications easier

Image credit: zhuweiyi49 / Getty Images

Snowflake has always been about storing large amounts of unstructured data in the cloud. His two recent acquisitions, Neeva and Streamlit, make it easier to find and build data-driven applications. Today, the company announced a new container service and his partnership with Nvidia to make it easier to harness all of your data and build generative AI applications to run on his Nvidia GPUs.

Snowflake Senior VP of Products Christian Kleinerman said the goal is to make data available to users without having to copy it and move it to an external application. “We want our customers to be able to bring computing to their enterprise data without having to send it to any kind of external system,” Kleinermann told TechCrunch.

The company introduced Snowpark Container Services and the ability to run containerized applications on Nvidia GPUs without moving data outside of Snowflake.

“We provide the ability for both customers and partners to run Docker containers within Snowflake’s security perimeter, providing controlled access to enterprise data residing in Snowflake,” said Kleinerman. increase.

“And the way we enable this container service is by offering a wider range of instance flexibility than what Snowflake has traditionally provided. The biggest vector for is access to the GPU,” he said. Here he capitalizes on his partnership with Nvidia.

Manuvir Das, Nvidia’s VP of Enterprise Computing, said he sees Snowflake as a place where businesses can store critical data, build applications on top of that data, and run those applications on Nvidia GPUs. It is said that it will be. It’s a very powerful combination, especially when incorporating generative AI into the equation.

Combining Nvidia’s GPU power with the NeMo framework, he says, companies can get their data on Snowflake and start building sophisticated machine learning models based on their own data.

“That’s why this partnership is great because Snowflake owns all the data and is the first to have an execution engine to run different software with that data. , there are execution agents built by Nvidia for training, fine-tuning, reinforcement learning, and more,” Das said.

“And the integration that we’re doing together is bringing that agent into Snowflake’s platform and integrating it so that every Snowflake customer who has data can take advantage of some kind of first-class functionality. “Custom models are generated by the Snowflake platform, which allows us to use data to create models,” he said.

By putting customer data, the models they create with that data, and the applications they’re running to access those models all in one place, he helps keep that data safe and secure. He said it would be easier to manage. Nvidia technology makes everything faster.

Snowflake’s Snowpark Container Services are available in private beta starting today.