Nvidia launches Al foundry business: joins hands with Meta Llama 3.1 open source model to customize and deploy super models for customers.

"NVIDIA AI Foundry" foundry service

Nvidia said customers can use Meta's Llama 3.1 AI model, as well as Nvidia's software, computing and expertise, to customize "super models" for specific domains.

Customers can train these "supermodels" with proprietary data, synthetic data generated by Llama 3.1 405B and Nvidia Nemotron Reward models.

NVIDIA AI Foundry is powered by the NVIDIA DGX ™ Cloud AI platform, which is co-designed with the world's leading public cloud to provide enterprises with vast computing resources that can be easily scaled as AI needs change.

NVIDIA AI Foundry includes NVIDIA-created AI models such as Nemotron and Edify, popular open base models, NVIDIA NeMo ™ software for custom models, and dedicated capacity on NVIDIA DGX ™ Cloud (built and powered by NVIDIA AI experts).

The output is taken care of by NVIDIA NIM, an inference microservice that includes custom models, optimization engines, and standard APIs, and can be deployed anywhere.

NVIDIA NIM Inference Microservices

IT House Note: NVIDIA NIM is a suite of accelerated inference microservices that allow enterprises to run AI models on NVIDIA GPUs anywhere, including cloud computing, data centers, workstations, and PCs.

By using industry-standard APIs, developers can deploy AI models with NIM in just a few lines of code. NIM containers seamlessly integrate with the Kubernetes (K8s) ecosystem and can efficiently coordinate and manage containerized AI applications.