Data-driven estimation of input–output relationships in chemical plants
About Me
My name is Kohei Morimoto, and I am currently an engineer at PFN. I participated in the PFN internship in 2023 and joined the company as a new graduate in 2025. I studied at Kyoto University, where I conducted research in control theory. More specifically, I worked on distribution control problems for stochastic systems—focusing on controlling probability distributions, rather than individual system states, toward a desired target distribution.
During my internship, I was assigned to the Plant Automation team, and I continue to work with the same team today. The team works on projects aimed at achieving autonomous plant operations by combining machine learning with control engineering know-how, particularly for complex plants where automation is difficult using conventional methods alone.
What I Worked On During the Internship
I worked on a project titled “Data-driven estimation of input–output relationships in chemical plants.” In system identification, ideal data is typically obtained by applying random inputs, such as white noise or M-sequences, to a system. However, applying such random inputs to a plant in actual operation is difficult from both safety and production efficiency perspectives. On the other hand, it is also well known that performing system identification using data collected under controlled operation is academically challenging. In this internship project, instead of aiming for full model identification, I focused on identifying only the sign of the step response, and attempted to address the problem using methods from causal inference.
Why I Applied for the Internship
The main reason I applied was that I had often heard PFN was an environment with many highly skilled and intellectually curious people, and I wanted to work in such an environment myself. In addition, since I specialized in control theory at university, I was particularly interested in the theme of control using machine learning.
What My Days Looked Like During the Internship
I typically arrived at the office around 9:00 and had a daily meeting with my mentor. For lunch, I often joined team members, where we had the chance to talk about both technical topics and casual matters. My main work involved reviewing papers, designing experiments, and conducting experiments and evaluations on internal clusters. I also had multiple opportunities to present my results to a broader audience, which included preparing presentation materials. There were also many opportunities to interact with other interns, such as having meals together.
What I Gained from the Internship
Since the internship lasted less than two months, I learned how to produce results within a limited timeframe. I also had many opportunities to receive feedback from team members, which helped me further develop my own research. At the time, I was living in Kyoto, so gaining a concrete sense of what it is like to work in Tokyo through the internship was also a valuable experience that later influenced my career decisions.
Who Should Apply?
From what I hear from my juniors, many people feel that PFN’s internship has a high barrier to entry. However, the screening process evaluates not only past achievements, but also how you approach challenges, your motivation, and how well you match with the theme. If you find the program even slightly interesting or feel that you can make use of your expertise, I would encourage you to take on the challenge and apply.
