Arricto The mission is to enable data scientists to build and deploy their machine learning models faster. The company that raised $10 million Series A round at the end of 2020builds its platform on the basis Cubeflow, an open-source cloud-based project for building machine learning operations that was originally developed by Google but is now largely community driven. So far, Arrikto’s main product has been a self-managed enterprise distribution of Kubeflow for enterprises (aptly named “Enterprise Kubeflow”) that want to run it in their datacenters or virtual private clouds. Today, the company is also launching a fully managed version of Kubeflow.
“Moving ML models from experimentation to production is incredibly challenging,” Arrikto CEO and co-founder Konstantinos Venetsanopoulos told me. “We see several common reasons for this. First, data scientists are not, in fact, operations experts, and operations specialists are not data scientists, and they do not want to become data scientists. Secondly, over the past couple of years, we have seen a rapid growth of machine learning tools. They are extremely fragmented and require a lot of integration. We see people struggling to sew everything together. Both of these factors create a huge barrier to entry.”
With a fully managed Kubeflow, Arrikto aims to provide businesses with a platform that can help them accelerate machine learning pipelines and free data scientists from having to worry about infrastructure while also allowing them to continue using tools they are already familiar with (think notebooks, TensorFlow, PyTorch, Hugging Face, etc.). “We want to break down the technical barrier that prevents most companies from deploying real machine learning capabilities,” Venetsanopoulos said.
The company claims that with Kubeflow as a Service, data scientists will have instant access to the end-to-end MLops platform. It’s basically Arrikto’s Enterprise Kubeflow, with lots of custom automation tools on top of it to abstract away all the details of the Kubernetes platform it’s on.
For now, Arrikto will only run on one cloud, but the plan is to support three major cloud providers in the long term to ensure low latency (and reduce the need to move large amounts of data between clouds).
Interestingly, Ventzanopoulos argues that the company’s biggest competitor right now is not other managed services like SageMaker AWS, but companies trying to build their own platforms by bundling open source tools.
“Kubeflow as a service provides data scientists and DevOps engineers with the easiest way to use the MLOps platform on Kubernetes without having to request any infrastructure from their IT departments,” said Venetsanopoulos. “When an organization deploys Kubeflow in production—on premises or in the cloud—Kubeflow as a Service from Arrikto speeds up the process.”
The company, which now has about 60 employees, will continue to offer Kubeflow Enterprise in addition to this new fully managed service.
Credit: techcrunch.com /