Nornickel develops new approach to secure AI use in the cloud with Yandex Technologies

Nornickel has launched a large-scale program to develop and deploy an ecosystem of complex AI agents across more than 30 key production and corporate processes. Yandex's cloud infrastructure provides the necessary computing capacity and enterprise-grade data protection.

Nornickel applies artificial intelligence across all processing stages and workflows - from mining and production to support functions - delivering consistent economic impact. The current development phase focuses on generative AI to improve the quality of management processes in both industrial operations and office departments.

To implement the project, the company built a hybrid architecture combining its own computing infrastructure with Yandex Cloud solutions (part of Yandex B2B Tech, the Yandex business

that develops enterprise solutions). Nornickel became the first major mining and metallurgical company to integrate large language models hosted in a cloud environment into its own corporate perimeter. As part of the joint project, the partners developed a new approach to the secure use of models in the cloud - specifically, connection via a dedicated physical link with no data stored on the cloud provider's side. This enables AI agents to operate within corporate networks without internet access, meeting information security requirements.

"We take a pragmatic approach to AI adoption. Our focus today is on scenarios where technology delivers measurable economic results or dramatically accelerates business-critical processes. Building our own competencies in domain models and complex agentic systems allows us to create a secure, vertically integrated ecosystem aimed at improving efficiency - especially in the context of talent shortages. Through this partnership, we are building a resilient technology model that simultaneously reduces the cost of maintaining solutions and ensures secure handling of sensitive data through our own computing cluster," said Vitaly Busko, Vice President for Innovation at Nornickel.

At the current stage, the company is independently developing complex agentic systems based on MetalGPT, a metallurgical domain model, to transform design processes, capital project management, inventory management, and financial and economic control. Large language models from the cloud are being integrated into existing business processes as digital assistants - primarily in the form of chatbots. Dozens of complex agentic systems are already in the testing phase, alongside a significant number of simple personal assistants for employees. In some scenarios, task completion time has been reduced from 30 days to just a few hours.

As part of the project, the partners are developing the cloud platform by testing it in real industrial use cases and

the tools for AI agent development with industry-specific requirements in mind. The next phase of the collaboration will expand access to AI tools for office employees. The companies are jointly developing a module for the Yandex AI Studio platform that will allow users without specialized training to independently create simple agents and address personal productivity tasks.

"Our partnership with Nornickel allowed us to deploy Yandex AI Studio in a real industrial environment with demanding requirements for security, scalability, and speed of implementation. In effect, we are setting the standard for AI agent use in the industry. Together, we developed a new approach to using large language models in the cloud - one that is already being adopted by clients in other sectors, including a top-10 bank and a major mining company," commented Grigory Atrepyev, CEO of Yandex Cloud.