Company Updates

Democratizing GPU Access for MLOps: Domino Expands Support to NVIDIA AI Enterprise

Thomas Robinson2021-08-24 | 5 min read

Return to blog home

The trend from our customers is clear: GPU-based training is a competitive advantage for enterprises building increasingly sophisticated models. We see this in everything fromNLP for predictive supply chains to image processing for histopathology in clinical trials, andpredictive maintenance for energy infrastructure to automatic claims resolution in insurance.

Domino partnered with NVIDIA in 2020 to certify the Domino Enterprise MLOps Platform on NVIDIA DGX systems. That support has allowed our customers to expand their usage of powerful NVIDIA A100 Tensor Core GPUs for AI and ML workloads. Customers can use Domino to train models on one to eight A100 GPUs in a single NVIDIA DGX system. And earlier this year, we announced support for multi-node training via Ray & Dask (in addition to Spark). This functionality allows for massive model training on large-scale DGX cluster infrastructure. Additionally, Domino’s support for the NVIDIA Multi-Instance GPU (MIG) feature in the A100 GPU helps customers easily deploy up to 56 notebooks or inference jobs simultaneously on a single DGX system.

But NVIDIA’s AI offerings extend beyond just DGX with NVIDIA AI Enterprise, built on theNVIDIA EGX™ platform. This architecture supports rapid deployment, management, and scaling of AI workloads in a variety of deployments, including on mainstream accelerated servers and the modern hybrid cloud.

That’s why today, we’re excited to announce that we’re working closely with NVIDIA to deepen our product integrations by enabling the Domino Enterprise MLOps platform to run with a broader range of NVIDIA GPUs, and validating it for NVIDIA AI Enterprise so that Domino runs seamlessly onNVIDIA-Certified Systems™.

The Difference for Customers is Clear

For Domino customers, the benefit is clear. By supporting the NVIDIA AI Enterprise software suite, Domino will expand our support for more NVIDIA-accelerated computing platforms. In addition to DGX, Domino will support a wide range of NVIDIA GPUs with mainstream accelerated servers, including the A100, A30, A40, A10 and T4, via packaging from Dell, HPE, Lenovo, or other hardware providers. This will provide customers the same certified Domino functionality across a broader range of NVIDIA-accelerated systems.

Since NVIDIA AI Enterprise enables hundreds of thousands of companies running VMware vSphere to virtualize AI workloads on mainstream NVIDIA-Certified Systems, this new integration makes it even easier for enterprises to procure, run, and scale Domino’s Enterprise MLOps Platform.

With expanded Domino support across more NVIDIA-powered platforms, enterprises get the best of both worlds: leading NVIDIA AI training and inference performance across a variety of systems, with all the necessary data privacy, integrity, and reliability.

Domino’s powerful, scalable, and collaborative Enterprise MLOps Platform helps data science teams leverage NVIDIA’s advanced GPUs as they build models for AI and machine learning solutions.

Building on a Record of Success

Domino’s powerful, scalable, and collaborative Enterprise MLOps Platform helps data science teams leverage NVIDIA’s advanced GPUs as they build models for AI and machine learning solutions.

Lockheed Martin has unlocked more than $20m in data science efficiencies and cost-savings across 300 data scientists by centralizing tooling across the enterprise with Domino, using NVIDIA as its choice in high-performance computing. Using the combination of Domino and NVIDIA, it streamlined collaboration and knowledge sharing while automating manual DevOps tasks that had hindered data scientist’s productivity.

Johnson & Johnson is also leveraging a combination of technologies from Domino and NVIDIA, building an Enterprise MLOps strategy to achieve model velocity. At the most recent NVIDIA GTC, our CEO Nick Elprin, along with Johnson & Johnson Enterprise CIO Jim Swanson, discussed the company's experiences embedding data science throughout the business, taking into consideration people, process, and technology.

AES, in less than two years, used Domino's Enterprise MLOps Platform to go from 0 to 50 models deployed in production. Utilizing NVIDIA GPUs, AES’ models cover a variety of different domains: predicting maintenance needs for power generation equipment, guiding fintech energy trades, making hydrology predictions, providing weather forecasting for utilities, and more.

We’re excited at the prospect of our growing partnership with NVIDIA, and what it is going to enable our customers to accomplish with models at the heart of their business.

Thomas Robinson is the VP of Strategic Partnerships and Corporate development at Domino, where he's responsible for building Domino's partner ecosystem, developing offerings providing differentiated value to partners. He previously acted as Domino's chief people officer, responsible for building an organization to unleash data science to address the world's most important challenges. Prior to Domino, Thomas worked at Bridgewater Associates, driving strategic transformation efforts, first as a director in Bridgewater's Core Technology Department to define the next generation of enterprise architecture, and then as a general manager focused on recruiting and retaining technical talent.


Subscribe to the Domino Newsletter

Receive data science tips and tutorials from leading Data Science leaders, right to your inbox.


By submitting this form you agree to receive communications from Domino related to products and services in accordance with Domino's privacy policy and may opt-out at anytime.