Nvidia founder and CEO Jensen Huang and Snowflake CEO Sridhar Ramaswamy announced the expansion of the two companies’ partnership. The move aims to simplify and accelerate enterprise artificial intelligence (AI) adoption by combining their data and AI capabilities.

Addressing thousands of attendees at this week’s Snowflake Data Cloud Summit, Huang emphasized the transformative potential of AI for businesses.

AI gives every company an opportunity to turn its processes into a data flywheel, he said, joining a virtual fireside chat from Taipei. Companies need to “take all the most important processes they do, capture them in a data flywheel and turn that into the company’s AI to drive that flywheel even further.”

Baris Gultekin, head of AI at Snowflake, echoed Huang's sentiment, highlighting the importance of creating enterprise-grade genAI solutions that address organizations' common challenges.

He added one of Snowflake’s focus areas is reducing hallucinations. “We reduce hallucinations by building a full stack solution, understanding when the LLM [large language model] should respond or should not respond to the question.”

The other focus is data governance. “How do you make sure that only people who have access to the data is able to use and see that data?”

Gultekin told SDxCentral that many organizations are building custom AI solutions, such as HR chatbots, to provide specific answers based on the user’s role. “So providing enterprise-grade controls around how to use AI is a big focus area,” he said.

Snowflake and Nvidia partner for generative AI The two companies integrate Snowflake Cortex AI and Nvidia NeMo Retriever and Triton Inference Server, designed to provide highly accurate results with retrieval-augmented generation (RAG) and expand the ability to deploy, run, and scale AI inference for any application on any platform.

Announced at last year’s Snowday event, Snowflake Cortex is a fully managed service designed to enable organizations to discover, analyze, and build AI apps in the Data Cloud more easily by bringing industry-leading large language models such as Meta AI’s Llama 2 mode, task-specific and advanced vector search functionality directly to the users.

At this week’s Data Cloud Summit, Snowflake introduced new features to Cortex AI including Cortex Analyst, Cortex Search and LLM fine-tuning.

In addition, Snowflake Arctic, the open and enterprise-grade LLM, is available as an Nvidia NIM inference microservice. With its mixture-of-experts (MoE) architecture, the Snowflake Arctic is optimized for complex enterprise workloads to offer efficiency at scale.

Nvidia NIM inference microservices is a set of pre-built AI containers and part of NVIDIA AI Enterprise, which  enables organizations to easily deploy a series of foundation models right within Snowflake.

These moves are built on Snowflake and Nvidia’s existing partnership announced last year. The data lake provider leverages the chip company’s GPUs, particularly for machine learning (ML) workloads, and to bring Nvidia AI Enterprise and NeMo LLM framework to the Snowflake Data Cloud.

Nvidia AI Enterprise, which is the software pillar of the vendor’s AI platform, includes over 100 frameworks, pretrained models, and development tools, including PyTorch for training, Nvidia RAPIDS for data science, and Nvidia Triton Inference Server for production AI deployments. The integration with Snowflake’s Data Cloud will expand Snowflake’s AI and ML capabilities, executives claim.

Democratizing AI Generative AI has parallel use cases across industries, Gultekin said.

For example, Bayer, a Snowflake customer, is using the Cortex Analyst product to streamline sales operations. This tool allows Bayer's sales team to query complex datasets easily and helps generate the right SQL to run and provides accurate answers, Gultekin explained.

“Being able to talk to that data is brand new and super powerful,” he added. “Democratizing the data and being able to glean even more insights from that data is what these products allow customers to do.”

Kari Briski, VP of AI software, Models, and Services at Nvidia, highlighted the democratization of AI as a key benefit of the partnership.

“Meet the real value of the democratization of generative AI is about choice and not being locked in. And so this is giving Snowflake customers the choice to find the models that are best for them and their data,” she told SDxCentral.

Generative AI future directions Looking ahead, Gultekin and Briski discussed emerging AI trends and future directions. They pointed out that the next phase of AI development involves enabling AI to take action, not just answer questions.

“Being able to do kind of introspection, reasoning, being able to call out to tools, being able to do things with that data with the systems that you've built is the next focus.,” he said. “I imagine next year is going to be all about agents.”

“It's like you asked a question it perceives it and then understands and can go take action,” Briski said. “Like last year, everyone was an exploration like what is generative AI. Open models really helped fuel that curiosity and exploration,  and then it's really now under production, and now that people are going into production, like wow, I needed to do so much more. It's already been great. But I want to do even more. I want to take action. So that's definitely the trend.”

The first action that most people are working on is SQL generation, she added.

Image: Nvidia founder and CEO Jensen Huang and Snowflake CEO Sridhar Ramaswamy. Credit: Snowflake.