January 11, 2023

Aiming to earn a doctorate at a graduate school in Kyoto while belonging to a Tokyo office

Working at Secure System Platform Research Laboratories
Junki Mori

Mr. Mori majored in research on the theme of quantum theory during his university years. After joining NEC, he changed his research theme to machine learning and has been engaging in research into federated learning. Since fiscal 2023, he has been working on earning his doctorate at a graduate school in Kyoto while remote working, utilizing NEC’s domestic graduate school study system.

Developing high-accuracy AI using data among organizations

The “federated learning” that I am working on now is a technology for machine learning that uses data held by multiple organizations. Normally, AI learning needs a massive amount of data. This has long created the problem of not having enough data to produce accuracy when companies, banks, hospitals, and other businesses tried to create AI. Federated learning is technology that caters to such issues and can generate high-accuracy AI models while protecting the confidentiality and privacy of data owned by multiple businesses.
This concept was originally advocated by platform providers to generate AI models while retaining the secrecy of the individuals’ terminal data. Drawing on previous business achievements, NEC is aiming to lead the world with data operation across enterprises and other organizations.
This is my second year on this research, which I started in my second year at NEC, and I am finally at the point of conducting a demonstration experiment with a customer and starting to feel new challenges upon entering a step where our research results are being useful in society.

Participating in international conferences even when starting out from a different background

I originally studied quantum theory in university. However, by the time I was finishing up my master’s program, I started wanting to do research that can produce research results that are directly useful to society. That is how I began looking for a place where I can do applied research on the theme of AI and machine learning. I knew that NEC often published research papers at renowned academic conferences, so I was interested in finding a job at NEC. I thought that being around top-notch researchers while publishing research papers would be a great opportunity for my growth as a researcher.
Quantum theory and machine learning are completely different areas of study, but I can still apply the same mathematical foundations such as linear algebra and calculus. After joining NEC, thanks to my supervisor giving me time to study and my senior colleagues teaching me programming from scratch, I was able to smoothly catch up.
In fact, I wrote three research papers in the three years of employment with NEC, of which two were accepted by an international academic conference. One of these was the International Joint Conference on Neural Networks (IJCNN), which is an international conference on AI. The conference was held in Venice this year, so I went on a business trip to the location. It was inspiring to have discussions with researchers around the world, and I also enjoyed sightseeing in Venice.

Studying for a doctorate utilizing NEC’s educational system

Having joined NEC, I feel that it has a fully-furnished environment for cultivating researchers. In particular, the domestic graduate school study system and the study abroad system are corporate systems that support your acquisition of a doctorate degree at a graduate school. The company pays for your tuition. I was selected for the domestic graduate school study this year, so I am planning to work on my doctorate at the graduate school of the university I graduated in Kyoto.
Now, to commute to Kyoto while belonging to a laboratory in Tokyo —NEC just rolled out the remote working system, so I am considering moving to Kyoto next year. The system is eligible for employees who can work remotely. The company will pay for up to one or two shinkansen (bullet train) trips a month to come to the office. My work is primarily done remotely, so I am considering using this system.
The reason for my studying a doctorate program is to make my own competitive strong point in the area of AI. Currently, I have found a new issue in federated learning concerning how to handle the difference in data distribution among organizations, so I would like to focus on that as my strength.

A day at work

Message to my past self in my school days