Element AI Launches Orkestrator AI-Powered Software to Manage GPU Clusters

Element AI announced Orkestrator, an addition to the company’s portfolio of AI-powered software solutions.

Designed for AI practitioners and the IT teams that support them, Orkestrator is a workload scheduling tool built as a result of the company’s need to optimize its own computing resources. A software-as-a-service (SaaS) version is available, with an on-premises version expected for Spring 2020.

Orkestrator helps customers future-proof their IT infrastructures through optimized utilization of GPU clusters, allowing AI practitioners to build quality models and remove most of the engineering heavy lifting required to efficiently use GPUs. By maximizing these specialized and in-demand resources, it will play a pivotal role in supporting the structural backbone of organizations deploying AI solutions for increased productivity.

“We originally created Element AI Orkestrator for our own internal data science teams and developers conducting AI research. The tool became so useful, our customers began requesting it,” said Ludwig Gamache, head, IT, Element AI. “Orkestrator is ideal for companies that need additional computing power to process AI solutions. Our own GPU usage is over 90% ; by using the Orkestrator tool we were able to grow our cluster by 36 times, yet required minimal additions to our IT staff.“

Orkestrator monitor
Click to enlarge

Using Kubernetes, Orkestrator provides AI researchers, applied research scientists, and developers with the computing infrastructure needed to build and maintain AI models. Providing transparency and control over job scheduling, it allows AI practitioners to conduct higher-level jobs such as batch processing and hyperparameter searching. IT administrators can simultaneously optimize cluster use, simplify node maintenance, and gain better visibility through dashboards.

“Orkestrator has a lot of potential for customers who have made significant investments in their AI infrastructures,” said John Barco, senior director, DGX software, Nvidia Corp. “It can help end-users make the most efficient use and scale of their computing resources.“

Click to enlarge

As organizations build and implement their own AI solutions that will need to be scaled over time-computing power and storage requirements will grow to accommodate heavier GPU-based workloads with increasingly complex resource distribution and management.

Video demonstration

About Element AI
It is in AI software that helps people work smarter. Founded in 2016 by serial entrepreneurs including JF Gagné and A.M.Turing Award recipient, Yoshua Bengio, PhD, the company turns research and industry expertise into software solutions that exponentially learn and improve. Its end-to-end offering, including advisory services, AI enablement tools and products, aims at helping large organizations operationalize AI and create business impact.The firm maintains a connection to academia through research collaborations and takes position in policy-making around the impact of technology on society.