Organizations that want to get serious about AI can only do so if their environment is ready for it. That is, it must be possible to build somewhat streamlined AI applications. And the AI must then also be accurate enough that it actually adds value. That is what DataStax intends to make possible with today’s announcement of its AI Platform.
Officially, the new offering is called DataStax AI Platform, built with Nvidia AI. That name doesn’t exactly chatter nicely, though no doubt several trademark gurus and legal experts will have been working on it for a while. More important than the name, however, is what the new offering will offer organizations. It should give enterprise organizations “a complete AI solution for all parts of the AI development and product lifecycle,” we read in a post from DataStax, from data ingestion and retrieval to application development and deployment to continuous AI training needed from the models that use those applications.
A single platform for more and more accurate AI
With this new platform, DataStax says it solves the two challenges we highlighted in the introduction. DataStax AI Platform, built with Nvidia AI is a unified platform. Today, it is still often the case that organizations use a host of separate tools. Those tools are often built for individual developers. That is not sustainable for organizations that want to seriously scale up. As a result, those tools often no longer work or don’t work well enough. By combining it all into a single platform, organizations no longer have to worry about this, is the idea.
Besides the possibilities that a unified platform like DataStax AI Platform, built with Nvidia AI offers in terms of scaling up AI deployment, there is of course still the concern about the quality of the AI that organizations deploy. This, too, should improve thanks to this new platform. Thanks to Nvidia NeMo Customizer and NeMo Evaluator, LLMs, SLMs and other models can be more easily trained. DataStax’s AI environment adds much-needed control over data search and retrieval. This is, of course, much needed to tailor GenAI to the various users who will be working with it.
Components of DataStax AI Platform, built with Nvidia AI
The goals from DataStax for the new AI Platform are clear. To achieve them, there are several components in the new offering. We briefly walk through them below.
From DataStax, there is the Langflow platform and DataStax Data Management. The former includes a development environment that focuses on creating and understanding complex logic as quickly and easily as possible. It does this using a visual interface. With DataStax Data Management, organizations can control everything around data management from one central place. Think of things like vector search, knowledge graphs, real-time AI analytics and so on. By the way, this is available in the cloud, as part of DataStax Astra, but also as cloud-native software in environments that customers manage themselves. Then it uses the DataStax Hyper-Converged Database.
From Nvidia, we see NeMo Evaluator and NeMo Customizer we already mentioned above, as well as NeMo Curator, NeMo Retriever and NeMo Guardrails. Also included in DataStax’s new AI Platform is Multimodal PDF Data Extraction, as well as NIM Agent Blueprints. These blueprints allow organizations to get up and running very quickly. There is a whole catalog of pretrained AI workflows that can be customized as needed. With this, developers immediately have a full suite of software, so to speak, that they can deploy at lightning speed to provide common applications with AI capabilities.
Like DataStax Data Management, the DataStax AI Platform, built with Nvidia AI is available in the cloud but also for environments that organizations manage themselves.
Also read: VAST further expands Data Platform: InsightEngine makes all RAG data available in real time