AI hardware and software provider Nvidia Corp. on Tuesday launched new NIM microservices designed to accelerate the development of AI apps.
NIM Agent Blueprints is a catalog of pre-trained, customizable AI workflows that enterprise developers can use to build and deploy generative AI applications.
The first three NIM Agent Blueprints presented are digital humans or avatars for customer service, multimodal PDF data extraction for enterprise retrieval augmented generation, and generative virtual screening for drug discovery.
NIM Blueprints give developers a head start in building applications with AI agents, Nvidia said.
The blueprints include sample applications built with Nvidia NeMo, Nvidia NIM, and partner microservices.
NIM Blueprints follow a GenAI market trend where many companies are trying to move from ideation to implementation.
Solving a GenAI problem
Although we are in the era of generative AI implementation, companies still struggle with challenges that prevent them from successfully deploying GenAI systems.
As a result, Nvidia NIM is trying to solve some of the problems companies face when implementing GenAI technology, including the speed at which companies integrate GenAI workflows, said Chirag Dekate, an analyst at Gartner.
“It’s about creating pluggable, modular components that companies can build on,” Dekate added.
While Nvidia could have chosen to provide developers with AI tools to build their applications, with NIMs like NIM Agent Blueprints, the vendor is providing companies with pre-trained models that help them move faster in building generative AI products and services, says Olivier Blanchard, an analyst at Futurum Group.
While this is cost-effective, it may not be what every company needs, Blanchard says.
“For companies that are completely unique, NIM may not be the right model,” he said. “They may start somewhere else and then use it to add additional layers of capabilities or features to their other product.”
What Nvidia is offering is not unique to GenAI technology, says Rob Enderle, an analyst at the Enderle Group.
“To be successful in the market, you have to have a complete solution, not just the technology,” said Enderle. “Initially, generative AI was a technology. Nvidia is now turning it into a complete solution.”
Nvidia isn’t the only vendor trying to make generative AI more than just a technology.
AMD, a competitor of Nvidia, is one of the leading companies in this movement, Enderle said.
The vendor last week acquired ZT Systems along with several other companies, including Mipsology, an AI software startup in France, to help with AI inference capabilities.
Google, AWS and Microsoft are also developing tools that make it easier for companies to easily deploy GenAI systems.
Nvidia Software Services
NIM Agent Blueprints enables Nvidia to extend its software services and make services such as NIMs available regardless of the cloud provider.
It’s about creating pluggable, modular components that companies can build on.
Chirag DekateAnalyst, Gartner
“This is part of the strategic plan that Nvidia is executing to essentially create a vertically integrated alternative regardless of where customers are developing their solutions at scale,” Dekate said. “By enabling these building blocks, like NIM Blueprints, Nvidia is essentially embedding higher-level modular architectures that basically run through the Nvidia stack regardless of where you run them.”
On Monday, Nvidia also introduced new CUDA libraries, including new LLM applications.
New LLM applications in CUDA include NeMo Curator and SDG or synthetic data generation.
NeMo Curator is an application that helps developers create custom datasets in LLM use cases.
Synthetic data generation is an LLM application that augments existing datasets with synthetic data to optimize models and LLM applications.
Esther Ajao is an editorial news writer and podcast host at TechTarget covering artificial intelligence software and systems.
