FSK Group is developing an AI platform called Kamerton, designed to create IT prototypes and automate document processing. IT Director Vladimir Kryltsov presented the system at the GoCloudTech 2026 conference.
The platform focuses on two areas: rapid prototyping for business hypothesis testing and the use of AI agents to automate repetitive tasks. One key application is the normalization of construction specifications, where the system standardizes data while flagging ambiguous entries for human review.
As of October 15, 2026, the infrastructure supported approximately 300 AI agents. Over the preceding 30 days, about 300 accounts accessed the system. Between early June and October 15, the platform processed around 42,600 requests and generated 19,000 output files.
The current architecture utilizes three computing circuits, including virtual machines, a task environment, and dedicated GPU resources. The configuration for 300 users consists of 16 CPUs, 96 GB of RAM, 2 TB of SSD storage, and a GPU with 16 GB of video memory.
Future development plans include upgrading the architecture to a unified runtime, centralizing request processing via a Request Core component, and enhancing security through stricter agent permission controls and integration monitoring.