A New Way to AI Training with GPUs

SwarmOne, the world’s first swarm computing company, has emerged from stealth mode after completing over 340,000 hours of AI training with a selection of first customers. The company, whose first executive team includes technology industry veteran and CEO, Harel Boren, and Israeli IDF Technology Unit veteran and CTO, Ben Boren, is set to revolutionize AI training by offering an instance-less software stack that dynamically and seamlessly performs training tasks across a global swarm of GPUs.

SwarmOne offers a different way of using GPUs for AI training. The technology behind the platform aggregates the AI model, training hyperparameters and data, abstracting away all the common MLOps, DevOps, SecOps and FinOps tasks, and then processing across multiple GPUs, which are situated in SwarmOne’s global network of certified data centers.

This means data scientists are free of the cloud, no longer having to set up and manage or optimize GPU instances. It also gives them access to massive compute power, all of which leads to better AI models, much quicker than with traditional providers.

On top of this, data scientists can train right from their development environment, with any training framework whether it be PyTorch, TensorFlow, HuggingFace, by adding just two lines of code. They will know the cost of training before they submit their training tasks, meaning that there are no more surprise costs from overruns.

Companies such as Salt Security – an AI driven firm and leader in the field of API Security - are already reaping the benefits of SwarmOne’s platform, with existing customers reporting an 84 percent time reduction before training even begins, a 91 percent improvement in delivery time, a 97 percent increase in AI quality and performance and a 67 percent reduction in compute costs.

“SwarmOne boosted personnel efficiency by about 90 percent, significantly reduced training costs and enhanced delivery,” commented Michael Erlihson PhD, AI Tech Lead at Salt Security. “This made us far more competitive in our market.”

“We built this platform because, like most data scientists, I spent way too much time sourcing, provisioning and managing GPU instances for [AI] training” said Ben Boren, CTO. “For AI to thrive, it must break free from cloud providers like Amazon Web Services (AWS), Google Cloud Platform (GCP) and Microsoft Azure. The barriers that exist in training are clear. But, it doesn’t have to be this way. Swarm computing can become the new standard for training.”

“The growth of AI, which is currently constrained by the huge amount of engineering cost and availability of GPUs, depends on a revolution in the way that training is done,” explained Harel Boren. “As the world’s first and only instance-less platform for AI training, we want to change the way that compute is used by humankind for good.”