When BakeryScan was launched in 2013, nobody could have predicted that this image recognition program would later save lives. In the mid-2000s, a Japanese bakery company, Andersen Bakery, discovered that the optimal business model was based on selling a large variety of unwrapped pastries. While signs could be used to identify products with similar appearances to customers, cashiers lacked this information. This made checkout a difficult and time consuming process. This led to long lines, drawn-out waits, and fewer customers. In order to reduce this, Andersen Bakery asked Hisashi Kambe to develop a way to automate the checkout process through the creation of BakeryScan. When creating this system, Kambe used an algorithmic pipeline, one of the leading AI methods of the time, which was carefully tuned to the problem at hand. And it worked. BakeryScan cut down on checkout times and operated with near-perfect accuracy. Four years later, a doctor watched a news segment on BakeryScan and had an idea: if this program could be used to differentiate between various pastries, maybe it could also be used to differentiate cancerous cells from healthy cells. The doctor reached out to Kambe and BakeryScan was adapted to the new purpose. Despite machine learning and neural networks dominating the AI conversation, the older algorithmic structure of BakeryScan allowed for it to be more adaptable than a newer machine learning program. Thus, AI-Scan was born. The developers were able to carefully tweak the parameters for new problems, which allowed AI-Scan to also be adapted for use on projects ranging from detecting flaws in jet-engines to analyzing art. That’s the beauty of technology. A tool can be created for use in something as specialized as pastry identification and it can then be developed in unforeseeable ways to change domains as far reaching as cancer diagnoses and electron lasers. BakeryScan also serves as a perfect example of the type of impact I would like to create in the world. I hope to use the intersection of my degrees in information science and physics to develop tools that can be applied to fields beyond my expertise. The research team that I am currently working on is in the field of particle physics, but I’ve noticed that much of the analysis work that I’m doing can be generalized to situations beyond those of neutrino interactions and kaon decays. I aspire to leave a legacy of well-written code and smartly designed programs that can stand the test of time. Just as AI-Scan’s algorithmic pipeline is still used today -- despite most image recognition using machine learning and neural networks -- I hope the contributions I make will outlast the tools I used. Machine learning algorithms are limited to only the problems which they were designed to solve, and when machine learning fades so will the algorithms. However, well-built solutions such as AI-Scan still remain relevant today, in spite of the AI conversation moving past the methods on which AI-Scan was built. As it stands now, machine learning provides great solutions to specialized problems; however, I hope to develop versatile solutions that have the potential to grow past the problem for which they were designed. Just as Kambe could have never imagined that BakeryScan would save lives, I hope my contributions to the world will improve it beyond the furthest reaches of my imagination.