Back to Careers Page

Adam Kern

Machine Learning Researcher
A photo of Adam Kern.
My biggest shift — and challenge — has been learning how to leverage my background in computer science for missile defense.

How has your career evolved at the Laboratory?

I joined the Laboratory less than two years ago as my first job out of graduate school, so the Laboratory has shaped my early career trajectory. My biggest shift — and challenge — has been learning how to leverage my background in computer science for missile defense. On the one hand, our group doesn’t need complex database systems or backend web servers, so I’ve put aside much of my software engineering knowledge. On the other hand, I’ve been applying my expertise in machine learning and artificial intelligence on a variety of projects, including a software that enables researchers to fit complex radar models to observed data and apply new deep learning techniques to the radar domain. I’ve been fortunate to have coworkers and leaders who have helped me make this transition and find where I can contribute.

Are you involved in any of the Laboratory’s outreach activities?

I’ve contributed to several STEM outreach programs, including Lincoln Coders, a program to bring the Girls Who Code curriculum to local classrooms; LL EduCATE, which aims to leverage staff expertise to provide technical curricula to middle and high schoolers; and an artificial intelligence/machine learning workshop at Hanscom Middle School.

The curious questions of new learners challenge me to look at concepts differently and reconsider my own understanding of technical material. For example, in one activity, students were building their own classification decision trees for objects from their classroom. One of them asked whether they could build a separate decision tree just for different types of pipe cleaners. It’s a simple question, but it raises some complex technical topics: how specific should the categories be, and who gets to decide? When do we stop splitting a decision tree to avoid overfitting a dataset? When do we need to employ hierarchical classification models? Looking through the lens of pipe cleaner classification brings a simple clarity to questions that can be technically quite messy or dry.

If you could meet anybody (current or historical), who would it be and why?

I would meet J. Kenji López-Alt. He’s a chef and cookbook author who has combined home cooking with the rigorous scientific methods you’d expect from an MIT grad. His love of both food and science really hits home with me. Over the past few years, I’ve come to really enjoy cooking because it provides me with an opportunity to practice a physical craft — something I don’t get to do often as a computer scientist. I’d love to pick Kenji’s brain on how his time at MIT has influenced his career as a chef, and thank him for the amazing resources he’s provided to the cooking community over the years. If you want to see his work, check out one of his articles on Serious Eats or pick up his cookbook, "The Food Lab: Better Home Cooking Through Science."