Everstar, a 17-person New York company founded in 2024, is tackling one of the biggest hurdles in nuclear energy: the paperwork. The bottleneck isn't the reactor technology itself, but the massive documentation required for licensing and compliance. Everstar's platform, Gordian, uses AI to streamline this process, reducing the time and cost of building nuclear reactors by an order of magnitude.
Gordian ingests decades of regulatory documents, design specifications, and engineering records, then uses AI models like Anthropic Claude and Azure OpenAI to generate defensible output for regulators. The platform runs on Azure, leveraging services like Azure Kubernetes Service, Azure Functions, and Azure AI Search. Everstar chose Claude for its superior reasoning and tool usage, particularly in long-horizon tasks. The company also relies on Microsoft Foundry to manage model consumption and compliance.
In a demonstration with the U.S. Department of Energy, Gordian generated a 208-page license application in just one day, compared to the usual four-to-six-week timeline. The AI also identified missing information and produced output that was reviewed and validated by human experts. Everstar's mission is to make nuclear energy more accessible by reducing the regulatory burden, and their technology is already showing promising results.