The database company DataStax Inc. is teaming up with Nvidia Corp. as it strives to become the data platform of choice for enterprises’ artificial intelligence initiatives.

In an announcement today, the company said it’s integrating its AI capabilities with the Nvidia AI Enterprise platform. The company claims the new integrated tools offering, dubbed the “DataStax AI Platform, Built with Nvidia AI,” can reduce development time of AI applications that leverage proprietary data by up to 60% in some cases. It provides everything developers need to fine-tune their models and improve the accuracy of their responses.

DataStax said it’s offering a complete solution for AI that covers everything from data ingestion and retrieval to application development and deployment, together with continuous training.

What’s in the stack?

The key components include DataStax’s Langflow platform, which provides an open-source visual framework for building retrieval-augmented generation or RAG applications. The DataStax Langflow platform was launched earlier this year, after DataStax acquired the creator of the open-source Langflow project, called Logspace.

DataStax also supplies its integrated Data Management tools, which encompass its flagship NoSQL database AstraDB with integrated vector search, hybrid search, knowledge graph, RAG, real-time analytics, streaming and other capabilities. DataStax became one of the first traditional database companies to add vector search functionality last year, enabling unstructured data to be stored as vector embeddings for easier retrieval by large language models.

With that update, it paved the way for DataStax’s RAGStack offering, which is an “out-of-the-box RAG solution.” RAG is a key technique used in AI development that makes it possible to provide additional context to LLMs from outside data sources. It allows models to deliver more accurate query responses, improving the performance of generative AI applications.

DataStax said AI demands extremely diverse kinds of data, and so an integrated platform that provides access to it all is preferable to bolting on different tools for vector search, knowledge graphs and so on.

Meanwhile, the Nvidia AI Enterprise platform adds a host of other interesting capabilities for AI developers, including Nvidia’s NeMo Retriever tool, which makes it easy to connect individual LLMs to very specific datasets, and NeMo Curator, a data curation tool for building large datasets for pre-training and fine-tuning models.

Other components provided by Nvidia include the NeMo Customizer, which is a performant and scalable microservice that helps simplify model fine-tuning and alignment for domain-specific applications. The NeMo Evaluator aids development by automating the evaluation process to test the accuracy of fine-tuned AI applications, while NeMo Guardrails makes it possible to add safeguards and prevent toxic or biased outputs.

Nvidia AI Enterprise also integrates multimodal PDF data extraction capabilities, providing a blueprint for ingesting unstructured data from PDF files, and NIM Agent Blueprints, which is a catalog of pre-trained and customizable AI workflows for creating and deploying AI applications.

The complete package

The DataStax AI Platform, Built with Nvidia, looks to be the complete package for AI developers, and companies will be hard-pressed to find a more comprehensive platform for building and deploying their AI models. Whether or not it’s the best platform of its kind remains to be seen, but DataStax is boosting its chances of success by making it as flexible as possible.

Enterprises can deploy the platform on any of the major public cloud platforms – Amazon Web Services, Microsoft Azure or Google Cloud – as well as on-premises environments, the company said. That last option makes it especially useful for enterprises in heavily regulated industries, such as insurance, finance and healthcare, the company said.

The integration makes sense because a lot of customers are using both platforms anyway, the company added. It explained that one of the problems enterprises face when bolting together various disparate tools for AI is that things have a habit of breaking down. For instance, the online travel agency Priceline.com LLC was already using DataStax’s AI offerings in combination with Nvidia’s NeMo tools, and it was spending a lot of time on trying to make everything work smoothly.

“It will greatly reduce AI development time,” said Priceline Chief Technology Officer Angela McArthur. “Having them integrated will greatly reduce the complexity for companies like us.”

Constellation Research Inc. analyst Holger Mueller said the integrated offering is interesting because it brings together Nvidia’s proven infrastructure with a reliable platform-as-a-service vendor in DataStax.

“The partnership makes it clear that Nvidia has ambitions in software too and it will help the company in that regard,” the analyst said. “It makes it much easier for joint customers to feed their data into Nvidia’s software and hardware and get their generative AI apps up and running. Some companies might be concerned about the dependencies they’re entering through this partnership, but most won’t worry as they just want to build their first, AI-powered applications.”

DataStax says the integrated platform will also provide more accuracy, giving developers more dynamic control over the data they feed into each AI application so they can improve their responses.

That’s especially important because companies are increasingly trying to use generative AI to improve productivity, with things such as PDF-driven chatbots for customer service and AI-powered analytics tools for surfacing business insights.

“The companies we’re talking to see these use cases as laying the groundwork for what they really want to do,” said DataStax Chief Executive Chet Kapoor. “They want to build ‘transformational’ AI projects that fundamentally transform how they operate and optimize for their customers.”

Image: SiliconANGLE/Microsoft Designer