From changing operating systems on school laptops to developing the models behind human behavior simulation, Milicevic followed a path shaped by curiosity, independence, and decisive choices.

At nine years old, Mateja Milicevic was already installing different operating systems on school laptops. By 12, he was writing code. His interest in computers developed alongside a strong affinity for mathematics, eventually leading him toward data science, predictive modeling, and a role at the frontier of human behavior simulation.

“I always liked figuring out how things worked,” Milicevic says. “Computers gave me a way to experiment with that from a young age.”

That curiosity became more consequential when Milicevic moved to the United States at 16 to attend boarding school. Leaving home meant separating from his family during a period when he was still learning how to manage daily life on his own. He missed important family events and had to adapt to a new environment without the support system he had known before.

“Moving that young teaches you independence very quickly,” he says. “A lot of things you would normally learn with your family around, you have to learn by yourself.”

The adjustment involved quieter challenges that were harder to identify than the practical demands of living in another country. Milicevic describes assimilation as subtler than people often expect, shaped by small differences that become noticeable over time. That experience influenced how he approached opportunities in education and work: it made him comfortable relying on his own judgment.

Milicevic came to believe that excessive hesitation keeps people from pursuing meaningful goals. He does not describe that lesson as permission to act carelessly. In his view, thoughtful decisions still require a point when consideration ends and action begins.

“You can think about an opportunity for a long time and still never feel completely ready,” he says. “Eventually, you have to decide whether you are going to try.”

Milicevic earned a bachelor’s degree in data science at Michigan State University, where his interest in predictive models sharpened. He was drawn to the challenge of using available information to estimate what might happen next. During college, he also founded Imagine Software, a student organization that grew into an educational community reaching more than 500 students — an early demonstration of his ability to turn a technical idea into something other people could use.

He continued into doctoral study in metagenomics at Michigan State, working on complex modeling questions at the intersection of biology and data science. Then Aaru appeared with a problem he found impossible to ignore: simulating human behavior at scale. The challenge aligned directly with his long-standing interest in using mathematics to understand outcomes, and the chance to help build frontier technology at an early-stage company outweighed the certainty of a finished degree. He left the PhD program and joined Aaru as the company’s ninth employee and one of its first three researchers.

“Aaru was working on a problem that immediately interested me,” Milicevic says. “Trying to understand and predict behavior at scale connected directly with the kind of modeling I already cared about.”

At Aaru, Milicevic took responsibility for developing and improving the mathematical models that power the company’s human behavior simulations. It is work closely connected to his earliest interests — and it carries the same demand for independent judgment he first learned at 16.

His fascination with artificial intelligence comes in part from its mathematical foundation. Milicevic points out that AI systems are built through enormous amounts of numerical calculation, yet their results can exceed what anyone would expect from the underlying operations. That relationship between mathematical structure and unexpected output continues to hold his attention.

“It is fascinating that artificial intelligence comes from numbers being multiplied,” he says. “The process is mathematical, and the result can still surprise you.”

His current work requires him to move between theory and practical model development. Human behavior is difficult to simulate, and useful systems demand sustained testing. The lesson Milicevic emphasizes most is persistence: people often overestimate the advantage of intelligence and underestimate the amount of work required to produce meaningful results. Difficult problems eventually test whether a researcher is willing to keep working after an early approach falls short.

“No one is so smart that the difficult work disappears,” Milicevic says. “You still have to keep going when the first idea fails.”

Milicevic wants to keep advancing simulation research. His interest extends beyond human behavior to simulation more broadly, with the larger goal of making complicated systems easier to examine and understand. The scale of the problems has changed since he first experimented with school laptops, but his attention remains fixed on the same question: what may happen next, and how mathematical systems can make that future easier to see.

“Curiosity is what got me here,” he says. “I still want to understand what is happening underneath, even though the questions are much harder now.”