"Academia is an individual pursuit and your research is your brand"

Myrl Marmarelis received his Ph.D. in Computer Science in August 2024, he was supervised by Greg Ver Steeg, Aram Galstyan, and Fred Morstatter. He was born and raised in Irvine, California, but lived in Greece briefly and attended elementary school there before returning to Palos Verdes, California for high school. He went to USC for his undergraduate degree, and his dad is a professor at USC.
Why did you choose computer science as a field?
As a kid, I stumbled into programming because I wanted to build websites. I learned about HTML and C++ to make a game with 3D graphics. While studying computer science at USC, I came across some research by Roger Ghanem, and ended up publishing my first paper with him on optimizing the 3D printing process. I always liked machine learning (ML) too. During one high school science fair, I tried to build an exploratory neural network. One ML idea I enjoyed was training algorithms based on evolution theory, by pitting different models against each other and letting the “fitter” ones continue evolving.
What did you choose ISI?
When I was applying to Ph.D. programs, I got a call from Aram Galstyan, Director of the Artificial Intelligence Division at ISI at the time. He invited me to come visit ISI. After the visit, I got accepted and selected for a fellowship, which was a surprise! I also got introduced to Greg Ver Steeg and was absolutely blown away by his creative ideas on information theory. I like how ISI allows lab members to have a lot of autonomy. We also have an abundance of resources, much more than a typical research lab. This setup lets us collaborate across many different disciplines and share our expertise.
What were your primary research focuses?
I focused on a field in machine learning called causal inference, where we draw causal relations and conclusions based on data. I took predictive models and developed them to become more robust and accurate.
What's the most impactful project that you worked on?
In the field of causal inference, you often have a treatment variable and data, and you want to find the difference in outcomes between the treatment and control groups. I worked on a paper that investigated what happens when the treatment variable gets more complicated, and how we can nudge treatments to improve outcomes. One example of this is in the public health space, where we can find out what dietary factors can be changed so people get healthier. There’s many confounding variables like exercise, pollution, and more that can affect health, so that’s where causal inference can come in.
What advice would you give someone who is applying to ISI for their Ph.D.?
Academia is an individual pursuit and your research is your brand. When reviewers are harsh or dismissive, know that this will push you to be a better researcher because you’re actively trying things out.
What are your plans after ISI?
I’ll be doing a postdoc at Caltech. In the long term, I want to pursue causal inference and other topics by seeing how they fit into industry, so this might translate into a startup or a research lab of my own. I’ve also been consistently working with ISI’s Abigail Horn on public health and dietary behavior research, so I hope to continue that line of research.