Building creative teams and scalable software. Currently working on bringing federated machine learning to the industry. We are early in a shift from a centralised cloud to a distributed cloud. Driven by the increased creation of datasets at the computational edge (think vehicles, Industrial IoT, drones, satellites), and barriers to data centralization (large data volumes, network limitations, security), we need software and infrastructure that lets us process data as close to where it is created as possible. For AI and machine learning this is a problem - today's machine learning pipelines tend to rely heavily on data centralization and in-cluster computation (think data lakes) and it is non-trivial to accomodate edge compute in machine learning operations. This is the core problem we set out to solve in Scaleout. Federated learning is a novel technology that enables training of machine learning models across geographically distributed edge nodes / clouds / devices. Training data is kept local - only parameters in a machine learning model needs to be sent to a server. We develop a software framework, FEDn, which lets developers integrate FL capabilities in their products and ML-pipelines, bridging the gap between Edge AI and contemporary machine learning operations.

Leverage distributed data for machine learning without the cost and risk of pooling data. Scaleout brings federated learning to MLOps. Our software platform lets you develop privacy-preserving solutions for computer vision, natural language processing, anomaly detection and more. Scaleout Studio is a cloud native platform for federated machine learning operations with two main goals: To 1) reduce the technical complexity of secure federated learning for the data scientist, and 2) to scale federated machine learning in production. Your team collaborates in Studio-projects to define the federated model and are then able to add and manage clients and devices in the federation. Studio is highly customisable, just add the components you need as Helm charts. We collect the leading open source tools for model development, collaboration and model serving, and provide them out of the box with project-level multi-tenancy and object level permissions. Studio runs on any standard Kubernetes cluster, and can be deployed in any public or private cloud and on on-premise hardware.