When Sierre Ternoey arrived in Aachen, Germany, she knew almost nothing about chemical engineering. But she had something else: a habit of diving into problems before she had all the answers.
Ternoey, an industrial engineering student at Northeastern University, was placed at RWTH Aachen University — Germany's largest technical school — for a research co-op. Her supervisor, Jan Pyschik, handed her what he called a "wildcard" project: a stubborn problem that other students had been warned to avoid because nobody knew if it could even be solved.
Chemical plants that make everything from cleaning products to clothing fibers rely on sophisticated software to design their operations. The software creates something called a flowsheet — essentially a blueprint showing how chemicals will be mixed, recycled, and separated inside a factory. The catch? These first drafts almost never work perfectly.
"What you get back is a long, dense report telling you something broke without telling you why," Ternoey explained. Engineers spend hours — sometimes days — sorting through the mess, using experience and, honestly, a lot of guesswork to find the problem.
Ternoey decided to build an AI that could do that job instead. She used skills from a first-year engineering course at Northeastern called Cornerstone of Engineering, taught by Professor Kathryn Schulte Grahame, known to students as KSG. That course throws students into big problems before they have all the tools, then teaches them to figure it out anyway.
"I think that it taught me to approach something that I'd never approached before and be comfortable with just starting," Ternoey said. She remembers thinking about a moment in KSG's class when she built her first electrical circuit — and then applying the same mindset in Aachen. "Okay, this is the same. We're going to start by making one flowsheet. We're going to watch one tutorial."
The results surprised everyone. In its best configuration, Ternoey's AI agent successfully fixed 26 out of 30 test cases — finding solutions without disturbing the engineer's original design choices. It could read the confusing error reports, diagnose what went wrong, and hand back a working simulation.
Pyschik admitted he was curious to see what she could achieve. "Honestly, I was a little nervous that I wasn't going to be able to cut it at my co-op," Ternoey said. But she found her footing — joining salsa lessons, training in taekwondo, and building a small community in a snowy, unfamiliar town.
Now, her tool could save chemical engineers worldwide countless hours of frustration. And it all started with a student brave enough to begin before she was ready.