Disclaimer: machine translated by DeepL which may contain errors.
After completing my degree in physics, I have been working as an industry researcher engaged in the research and development of medical AI. My goal is to create technologies that can help improve the working conditions of physicians and the management challenges currently faced by hospitals.
As a student, I spent my days immersed in research on bacteria. My core interest lay in “quantifying the unknown to gain understanding,” and from my fourth undergraduate year through my doctoral studies, I conducted biophysics research that aimed to understand biological phenomena through physics. Although bacteria seem far more straightforward than humans, once you begin observing them yourself, you encounter countless phenomena that defy simple explanation. In particular, I studied the collective behavior of bacteria: observing them under a microscope, quantifying features from the resulting images, and using physical theories and simulations to reproduce and understand what I saw.
I am now engaged in research on deep learning technologies for analyzing medical data, such as medical images and electronic health records. At first glance this may appear to be an abrupt shift, but I am still pursuing the same underlying theme—“quantifying and understanding the unknown”—this time with humans and AI as my subjects. More specifically, I work on quantifying the decision-making criteria used by deep learning models during image-based diagnosis and comparing them with the criteria used by physicians, in order to deepen our understanding of model behavior. While my job is ultimately to build AI systems, deploying AI that can withstand clinical use requires a deep understanding of how it behaves. In fact, many people with backgrounds in the fundamental sciences are active in this field both inside and outside the company, and the demand for scientists who apply scientific reasoning as scientists is clearly growing.
AI research itself is fascinating, but there is also a unique appeal to doing research within a Japanese Traditional Company (so-called JTC). Setting aside the notoriously long internal manuals, one major advantage is the company’s extensive connections with external organizations, a hallmark of its traditional nature. This has allowed me to work closely with medical professionals across a variety of institutions, broadening my perspective as a researcher. Within the company as well, there is a diverse community of researchers not only from the sciences and engineering but also from the humanities, creating an environment where it is easy to learn about fields beyond one’s own.
There were several reasons I chose to pursue a career in industry, but at the core was a simple curiosity: if possible, I wanted to experience research in many different worlds. Naturally, I had concerns about switching fields, but I embraced the spirit of Antonio Inoki’s famous phrase: “Take the first step, and that step becomes the path. Don’t hesitate—go forward, and you’ll understand.” So far, things have worked out. In recent years, terms such as “VUCA” and “BANI” have become widespread, emphasizing how uncertain and unpredictable our world has become. I feel that the need for the scientific ability to grasp the essence of the unknown is growing not only in AI research but across industry as a whole. For anyone unsure about their career path, taking a step into a field outside academia—and even into a field different from your own—may open up an entirely new scientific landscape.
Presentation on Astellas' translational research initiatives at the Immuno-Oncology 360° Conference (IO360, an international conference in the field of immuno-oncology)