Alumni Profile: Maxwell Crouse, IBM Research

Maxwell Crouse (PhD ’21) earned his PhD in computer science with a focus on artificial intelligence (AI) under the supervision of Ken Forbus, Walter P. Murphy Professor of Computer Science at Northwestern Engineering and professor of education in Northwestern’s School of Education and Social Policy. Crouse’s dissertation, “Question-Answering with Structural Analogy,” introduced an approach to question-answering systems that uses analogy to adapt an existing, general-purpose natural language converter — called the Companion NLU semantic parser — to answer questions without the need to learn from scratch for each new domain application. The method simplified the question-answering task, allowing for better performance and data efficiency.

Crouse joined Northwestern CS in 2015 directly following his undergraduate program in computer science at Indiana University Bloomington. After graduating, he worked at Microsoft as a software engineer on the Azure Storage team. He joined IBM Research as an AI research scientist in November 2021.

We asked him about his research interests, the lifelong skills he learned at Northwestern CS, and his advice for current students.

Why did you decide to pursue your PhD at Northwestern CS?

I was really interested in the cross-disciplinary nature of CS at Northwestern. My adviser, Ken Forbus, is working on AI techniques that draw on a lot of research from the cognitive science community, and ultimately I thought that research direction was a compelling path toward building smarter machines.

Could you describe your research interests in one sentence?

I'm primarily interested in the design of neuro-symbolic AI systems, such as systems that combine traditional symbolic AI methods with modern deep learning.

What skills or knowledge did you learn in the PhD program that you think will stay with you for a lifetime?

I'd say the most useful skill has been knowing how to get up to speed on a topic quickly. There's a surprising number of sub-skills that go into that which I think people often gloss over, such as recognizing when you can and can't trust a paper's results.

How have you used the skills you learned during the PhD program in your roles at Microsoft and IBM Research?

The work I did at Microsoft involved a significant amount of machine learning, so I ended up drawing a lot on the AI fundamentals I learned in the program. It was very interesting and certainly kept me on my toes.

Fortunately, I had the opportunity to shift into a research scientist role in the neuro-symbolic research group at IBM Research, so I get to use far more of the skills from my PhD in my day-to-day work. My work at IBM Research involves bringing together symbolic and deep learning techniques to solve problems in interpretable, effective ways, which means I must draw upon the research I did at Northwestern quite frequently.

What advice do you have for current Northwestern CS PhD students?

I'd tell them to keep in mind that progress isn't always linear and clearly defined. Students put an expectation on themselves that they should be producing X amount of work per year, like publishing one paper in a top conference per year. Certainly, one should strive to produce as much as they can, but they shouldn't measure their own success with generic rules-of-thumb. Someone can graduate with just one paper — but that paper is really impactful — or they might have three years with no papers and one year with three papers.

Many factors go into work output, such as individual characteristics, research area, and conferences targeted. A surprising — and unfortunate — number of those factors are out of the student's control. Focus on doing as much as you can and rely on your adviser to provide you with fair evaluations of your performance.