The success and growth of companies can be determined by the technologies they rely on in their tech stack. To deploy AI enabled applications to production, companies have discovered that they’ll need an army of developers, data engineers, DevOps practitioners and data scientists to manage Kubeflow — but do they really?
Much of the complexity involved in delivering data intensive products to production comes from the workflow between different organizational and technology silos. Integrating components manually requires a significant amount of resources, maintains the organizational silos and creates large technical debts.
Our approach to the challenge of getting AI enabled applications to production has been to embrace Kubeflow, adding it to our managed services catalog, and bridging the functionality gaps that exist. Nuclio and MLRun — our open-source frameworks — extend Kubeflow’s functionality by enabling small teams to build complex real-time data processing and model serving pipelines from the training results and data from the real-time feature store in a matter of minutes.
We believe that organizations should deploy a data science solution that breaks down silos between roles and abstracts away much of the complexity, while enabling high-performing, scalable and secure models that can be deployed in any cloud or on-prem.
Read on for our approach to extending Kubeflow’s functionality into a true end to end ML and MLOps solution.