Atomic Canyon wants to be Chatgpt for the nuclear industry

Technology companies Betting severely This nuclear energy can help provide the electricity they need to achieve artificial intelligence plans. But data centers need energy tomorrow, and the nuclear industry is not known for its speed.

Trey Lauderdale believes that artificial intelligence can give the nuclear speed you need.

Lauderdel’s obsession started with nuclear behavior near the house. In San Luis Obebiso, California, where he lives, he continued to run to people who worked at Diablo Canyon energy station. “They are like our football team coaches,” he said.

In talking to them, he learned that nuclear power plants swim in documents. He said Diablo Canyon, near Lauderdiel’s house in Saint Louis Obebaso, had about two billion pages. Lauderdiel, a serial health care man, had been to have mine that artificial intelligence could help the nuclear industry tame its paper problem.

Lauderdale was established The atomic valley A little more than a year and a half ago, he initially funded him with his own money. The startup of artificial intelligence is used to help engineers, maintenance technicians and compliance employees to find the documents they need.

The startup company obtained a deal with Diablo Canyon in late 2024. Lauderdale said the deal led to inquiries from other nuclear energy companies. “This is when I knew, as a businessman, we were at a point we needed to raise a round of capital.”

Atomic Canyon closed a $ 7 million seed tour led by Energy Impact, and the company also told Techcrunch. Among the participating investors are the commeal Ventures, Plug and PLAY VENTERES, Tower Research Ventures, Wischoff Ventures and former owners.

When Atomic Canyon started for the first time, artificial intelligence engineers tested various models with enormous results. “We have quickly realized hallucinations from artificial intelligence when you see these nuclear words.” “He didn’t see enough examples of shortcuts.”

But building a new model for Amnesty International requires a huge computing force. Lauderdel on his way spoke to a meeting with the Oak Ridge National Laboratory, which is taking place nuclear research and also happens to have the second fastest super computers in the world. The laboratory was fascinated by the idea and gave Atomic Canyon 20,000 GPU an hour.

Atomic Canyon models use a sentence, which is especially suitable for indexing documents. It aims to place a search station documents that can be searched using a recovery generation, or a rag. RAG uses large language models to create responses to the queries, but it requires LLMS to indicate specific documents in an attempt to reduce hallucinations.

Currently, Atomic Canyon is committed to searching for documents, partly due to the fact that the risks are less.

“One of the reasons why we start the obstetric work around the documents addresses is that obtaining this error may cause a person to be frustrated. No person is at risk in the factory,” said Luddeliel.

Ultimately, Lauderdale Canyon Self -Create Self -Create “First Round Dreaches” of documents, with references. “You will always have a person in the episode here,” he said.

Lauderdale has not set a schedule on this effort. He said that the research is the “founding class.” “You have to search.” In addition, given the number of documents in the nuclear industry, “We have a long runway in the search alone.”