UMD Team Wins Nuclear Energy Agency’s First Coding Competition

A multidisciplinary team from the University of Maryland won first place at the Nuclear Energy Agency’s (NEA) inaugural Coding Competition, beating 22 teams from 12 countries to create machine-readable risk registers using artificial intelligence algorithms. 

The team, led by Center for Risk and Reliability Director and Professor Katrina Groth, won a travel stipend to present its model at the NEA's International Workshop on Artificial Intelligence for Nuclear Energy next month in South Korea. 

“I am proud to see UMD innovators coming together to solve important problems in nuclear safety,” said Groth.  

Teams were given four weeks to use AI to convert risk registers from diverse human-readable formats, including Microsoft Excel, Word and PDF, into machine-readable data. 

The UMD team of Adrian Maker ’26 (chemical and biomolecular engineering), Stefano Marchetti (reliability engineering postdoctoral researcher), Cristian Schaad Ph.D. ’26 (reliability engineering) and Somil Varshney ’26 (computer science and finance) developed a method that uses large language models to transform complex, human-written risk registers into standardized digital data that can be analyzed automatically. 

For the nuclear industry, this is particularly valuable because safety and risk information is often stored in diverse document formats that are difficult to use consistently across organizations. 

The team’s method can help nuclear organizations better use the risk information they already produce, turning static documents into structured knowledge that can support analysis, trend identification and more consistent decision-making. 

“Our team wanted to show that AI can be used in the nuclear field in a way that is both innovative and trustworthy,” said Marchetti.