With advancements in machine learning technology, the complexity of the development environment has increased the importance of a scalable, standardized computing platform. Lablup is a platform-as-a-service (PaaS) solution that provides cloud and server management and scaling automation specific to machine learning environments that require a large amount of system resources.
Launched in November 2016, ‘lablup.ai (open source project SORNA)’ is a distributed-processing-based, open source execution platform that is responsible for smoothly distributing program code for design and training of deep learning models in server and cloud environments.
‘Lablup is based on Docker technology, which efficiently provides multiple independent operating environments on a physical server. Thus, it is possible to run arbitrary user code securely in a platform-agnostic, resource-constrained environment and to use various programming languages without any complicated configuration Support.
SORNA is available in two forms: open source and cloud backend hosting. Companies with physical servers can build and use open source SORNA on existing servers, while startups or small businesses without an IT infrastructure can be deployed in the cloud.
Lablup provides web-based, personalized programming education specialized in STEM field through CodeOnWeb, an online coding education platform developed based on the SORNA engine. Lablup is also developing a machine learning based natural language processing API along with a streaming interface that services various machine learning models in the form of API (application program interface).
The company plans to deploy SORNA through a two-track strategy: open source solution for companies with internal server farms, and cloud solutions to small and medium enterprises that lack IT infrastructures. As Lablup focuses on improving and optimizing the solution, it will continue to develop AI applied framework based on SORNA technology.
“As innovations made by machine learning and data-based systems have broadened the application range of IT environments, inevitably a limit to the scalability and convenience of the systems have surfaced” said CEO Jeong-Kyu Shin. He proposed that Lablup will support engineers to have the ubiquitous experience in complex computing environments.
“The design and training of a machine learning model requires a tremendous amount of computational resources, such as CPU and memory, which is a typical area where resource allocation issues are prominent,” said K-Cube Ventures partner Jun Kim. He expects Lablup to “work as a bridge to resolve the gap between machine learning experts and server management experts through team members’ specialized expertises in machine learning, big data, virtualization and distributed processing.”
“We expect Lablup to provide unique values to customers with high scalability and efficient resource consolidation management, such as the ability to run server farms directly using the open source version, and to use AWS, Microsoft Azure and Google Cloud services together when needed,” said Ji Seong Oh, senior investment manager at Stonebridge Venture Capital.
Inspired by Platum