Noticing my reflection in the ebbing water, I lift the seaweed covering the glistening tide pool. Peering inside, my eyes catch a delicate movement. I reach down, close my hand, and fish out a vibrant fistful of lush ulva, scarlet caulerpa, and sand full of fragmented shells. Sifting through this treasure in my cupped hands, I’m left with a cluster of minuscule hermit crabs. Antennae and legs slowly poking out of a few of the more daring ones,  I note their distinct, cerulean bands as they quickly scuttle back into the water, while the rest of the shy crabs are still scared frozen. 

Putting my childhood curiosity for marine life to use, I pursued research with the Zacherl oyster lab. Sitting in muddy crates before dawn, I found immense joy digging in the mud for live oysters, never knowing what else I’d find. Yet I was unsatisfied with our inefficient data collection methods and grew impatient not witnessing tangible results. 

There had to be a better way. Scouring alternative data collection techniques I deemed “still too inefficient”, I finally found it. While referencing pictures of our fieldwork, I remembered using Google Lens, a machine learning application to put names to those elusive creatures I encountered in my childhood tidepool adventures. Maybe we could automate the way we harvested data using these photos rather than tediously hand-collecting it, saving so much time for our groundbreaking researchers. 

Inspired, I proposed the idea of developing a machine learning model to Dr. Zacherl, who immediately became invested. Just like that, my simple idea matured into a sophisticated project. There was just one problem: nobody in the lab, my mentor included, had experience with this technology. I couldn’t even answer her seemingly basic questions like, “What is the optimal split for the dataset?” or “Can you use photos you intend to analyze in the training set?” Expectations were heavy on my shoulders and I was scared I’d make the wrong decisions.

Seeing my model fail to identify what was so obvious to my eyes felt like empty hands after combing through a promising tidepool; I was utterly confused. Struggling was familiar, but not having someone to look to for answers wasn’t. In water polo, I never had to worry about thinking for myself and just had to follow the coach’s play. In cello, I never had to come up with my own interpretation of a piece, I just had to recreate my teacher’s. I missed having decisions made for me, but hearing my mentor tell her peers about this incredible project, I couldn’t bear giving up on my vision if she could see it too.

I scoured the internet for every tutorial I could find, even resorting to ones that weren’t in English, and I rummaged through published research on machine learning used in diverse scientific fields like agriculture and neuroscience. After countless training iterations and augmentations to the image dataset, the model was finally strong enough. Not perfect, but strong, and for once, that was okay with me. Slowly but surely, I began to learn that science is defined by the pursuit for improvement, not perfection.

Now, when I present at scientific conferences and field questions from seas of brilliant naturalists, I know that my research is the work of an emerging marine ecologist with promising impact. I’ve previously struggled to make my own decisions, leaning on others to tell me what to do and faithfully carrying out what I was told, but here, my decision to actualize my vision is one I can truly call my own. 

Back in the tide pools, I stumble across another “daring” hermit crab confident in its ability to emerge from its shell to escape danger when no other crab is willing to. Smiling, I can’t help but think that I’ve grown to become like this little crustacean. When I trust myself, I can achieve so much more.