U.S. Nuclear Industry AI Assistant NIVA Opens After 6-Month Pilot

en.Wedoany.com Reported - U.S.-based Atomic Canyon has made its nuclear industry virtual assistant NIVA available to the commercial nuclear power sector in North America. The platform was developed by the company in collaboration with the Institute of Nuclear Power Operations, the Electric Power Research Institute, and the Nuclear Energy Institute, and member plants of these organizations are eligible to apply for access. This rollout means that platform access eligibility has been expanded to the industry level, and does not indicate that every nuclear power plant has completed local system deployment.

NIVA launched with two initial features. The Knowledge Assistant enables retrieval of nuclear regulatory guides, technical reports, and industry standards through natural language queries, with corresponding references cited in the responses. The Operating Experience Assistant searches historical operating events based on semantics and context, allowing staff to further analyze similar equipment issues and lessons learned. The Fault Diagnosis Assistant remains under development, with a targeted pilot planned for later in 2026.

The platform is built on the Neutron workbench and the FERMI retrieval model purpose-built for the nuclear industry. FERMI was developed using the Frontier supercomputer at Oak Ridge National Laboratory, with training data covering more than 53 million pages of U.S. nuclear regulatory documents. NIVA primarily connects to industry public and shared resources; for internal data such as plant licensing bases, operating procedures, maintenance records, and engineering drawings, separate configuration is required within the plant environment via Neutron Enterprise. The system is positioned as an information retrieval and knowledge assistance tool, and does not replace the judgment of nuclear professionals or directly participate in reactor control.

The two core features underwent a six-month pilot prior to release. Constellation Energy was among the early adopters, with nuclear assets including 21 reactors across 12 plants. The pilot focused on reducing the time required to locate technical documentation and historical operating records, supporting maintenance preparation, design change studies, and workforce knowledge transfer.