ARCS Oregon Scholar Derek Lilienthal knows “the field of AI moves quickly, and being able to generate and refine impactful ideas at speed is a crucial skill.”  He’s pushed himself in the first year of his PhD program examining Artificial Intelligence while simultaneously working part-time at the Pacific Northwest National Laboratory (PNNL), located in Richland, Washington. Lilienthal, an ARCS Scholar at Oregon State University, admits his dual roles “have been demanding. I often work long hours, but I genuinely enjoy both my research and my work at PNNL.”

His PhD research focuses on the privacy and security of large-language models (LLMs).

Lilienthal says his first year at OSU has required him to shift his mindset. “I come from a software engineering background, which is great at building systems.” Research, he says, requires a focus on identifying and framing deeper scientific problems. His future goal, he explains, “is to scientifically advance the safety and security of AI agents,” which will require a “researchers’ mindset to analyze problems, identify limitations in prior work, and iterate on ideas efficiently.”

At PNNL, he is a software engineer developing AI applications for various US government agencies. He collaborates with engineers and scientists to develop, train, and deploy generative AI models across high-performance computing clusters and cloud environments. While only some of his research overlaps with his role as a software engineer, Lilienthal hopes to transition to more research in the future at PNNL, one of the huge national labs in the US.   

Lilienthal began at PNNL in January 2024 in the High-Performance Computing Group (HPC), which is also known for working with supercomputers. He had previously completed AI-related projects during internships at other companies. When the opportunity arose to join PNNL and work with AI on supercomputers, he was eager to gain experience. He hopes to take the skills learned at PNNL and apply them to his research.

As AI has become more available to the general public, Lilienthal says, “What’s important to remember about AI is that, at its core, an AI model is just math. It’s easy to forget that something like ChatGPT is fundamentally a series of matrix multiplications and non-linear applications. While it’s undeniably impressive that an applied math problem can produce the kinds of capabilities we see in the news or experience firsthand, we must remember that AI is still a tool, not something with real intelligence.“

Lilienthal says the ARCS Oregon Scholar Award “greatly supported my wife and me by helping reimburse our relocation expenses” for moving to Oregon for his PhD and also “allowed us to pay down some credit card debt.”