Rubytech has integrated its proprietary AI infrastructure management platform, Spektr AI, into the Scalar AI Machine. This software-hardware complex is designed for developing, training, and running AI models, featuring equipment from the Ministry of Industry and Trade registry and software from the Ministry of Digital Development registry.
The integration of Spektr AI allows organizations to increase GPU utilization from a typical 30–40% to 70–85%. By managing GPU nodes as a unified cluster with a built-in scheduler, the system optimizes resource distribution among teams, reduces idle time, and delays the need for additional hardware purchases.
The platform operates on Deckhouse Kubernetes Platform and supports Astra Linux and RED OS. It provides centralized management, role-based access, and the ability to reserve capacity for critical tasks. The solution scales up to 32 nodes, with each node supporting 8 GPUs.
Key system components, including operating systems and container virtualization, hold FSTEC Russia certificates. The platform supports LLM inference, model cataloging, token counting, and security policy enforcement, enabling efficient resource control for data science teams and AI-driven business applications.
