Eliana Stefani

Eliana Stefani

What brought you to UC Santa Cruz?

At Cal Poly I got really interested in robotics and from there I worked in industry for about two years. I did some research in quantum computing and deep convolutional neural networks and from there, I recognized that there was a big gap in my knowledge and that while I was able to put together robots and put their trajectories together, there was a lot more that I could learn about how these robots think. Because of that, I applied to UC Santa Cruz so I could get a little bit more background in artificial intelligence and machine learning.

Why did you choose the Baskin School of Engineering for your graduate education?

What I like about the Baskin School of Engineering is the research being done in AI and machine learning and the collaboration happening between departments so they can help build off of each other in ways that they weren’t intuitively familiar with. Lastly, it’s absolutely beautiful being on a campus where you’re literally in the trees.

Tell us about your research

I’m doing research in unsupervised object learning which learns the existence of new objects given a set of data. This is currently an unsolved problem in machine learning and I hope my approach will eliminate the need for humans to label data and predominantly for humans to be able to recognize these types of objects. Since background objects are typically single, homogeneous regions and foreground objects are typically heterogeneous groups of regions, I’m using a combination of image segmentation, region clustering, and a learning heuristic to form a hypothesis on these learned objects. This line of research will advance both object recognition software as well as the computer’s ability to quickly pick out correlations in data that would have otherwise been indistinguishable to the human eye.

For example, humans are really good at learning a new object, say, visually. We think, “Oh, I’m walking down the street, I’ve never seen that thing before but I’ve seen it a couple of times now and I’m going to make a hypothesis: that’s a new species of animal that I’ve never seen before.” We’re very fluent at doing that visually, but with other data, like genomic data, we are very bad at that. We look at a human genome and think, “I don’t know what that is,” so my goal is to have a computer understand how humans can learn the existence of new objects, form hypotheses on these objects, and then in the future say, “I’ve seen that instance a couple of times before, I’m going to say that’s new object A or new object B.” If we can apply that to genomic data, I believe this will help us better recognize new patterns in genomic data to help us understand which combinations and patterns of genomes lead to certain results in terms of eye color or whether someone is predisposed to a certain disease or not.

How will your research help you at work?

I work at Lockheed Martin doing a little bit of machine learning research. While the exact type of research is not exactly the same as what I’m doing here at UC Santa Cruz, there are a lot of parallels, which is really helpful because what I learn here at school I can immediately implement in industry and that was one of the main motivators for me going back to school. Having a team at work has been very valuable in helping me progress past a lot of problems that I’ve been having in my research. So while it’s typically more common for people to go straight from undergraduate to graduate school, I do see huge value in being able to spend a little time in industry to be able to see those parallels between your work and your schooling.

What do you like to do in your free time?

I like mountain bike racing. I used to race for Specialized. I’m also slowly starting to get into SCUBA diving. When the conditions aren’t right for either of those, I love hiking at Big Basin.

Department: 
Electrical and Computer Engineering
Degree Program: 
Ph.D., Computer Engineering
Place of Birth: 
San Francisco, CA
Undergraduate Institution: 
Cal Poly
Graduate Institution: 
UC Santa Cruz
Advisor: 
Gabriel Elkaim