[Power K Woman] Soojung Ryu, Seoul National University Professor: "Contributing to Expanding the Ecosystem for Leading Processors in Korea"

Female Leader Who Led Saffion in Semiconductors
Experience at Leading Companies Like Samsung Electronics and SK Telecom
Active Role Bridging Academia and Industry
Researching Next-Generation Semiconductors Such as NPU and PIM
"Set Long-Term Goals... Embrace Diverse Challenges"

"If there is anything I can do to help establish a leading (semiconductor) processor company in Korea, I am willing to do it."

Soojung Ryu, Visiting Professor at Seoul National University, said this during an interview with Asia Economy at the Haedong Advanced Engineering Building on the Seoul National University Gwanak Campus on September 18. Professor Ryu is a prominent female leader in the semiconductor field, having led Saffion Korea (now Rebellions), a major domestic AI semiconductor company, until last year. She has played a significant role not only in next-generation semiconductor research, bridging academia and industry, but also in expanding related businesses and markets in Korea.

Despite earning her Ph.D. from one of the top engineering schools in the United States, Professor Ryu recalled that it was difficult to find a place to work in Korea that matched her expertise, as there were no leading companies in the processor field of system semiconductors. This is why she emphasized her determination to find her role in strengthening Korea's system semiconductor competitiveness. She has also focused on expanding and developing the overall domestic industrial ecosystem. The following is a Q&A with Professor Ryu.

Soojung Ryu, Visiting Professor at Seoul National University (former CEO of Sapeon Korea), is being interviewed by Asia Economy at the Haedong Advanced Engineering Building, Seoul National University, Gwanak-gu, Seoul.

―What prompted you to enter the semiconductor field?

▲Until my master's degree, I studied databases (DB). I graduated from a computer science department and later switched my major to computer engineering. In my senior year as an undergraduate, a professor specializing in databases invited me to study together, so I pursued my master's degree. Although I conducted research related to databases at the time, I was also very interested in computer architecture. Later, I went to the United States for my Ph.D. and met professors in both the DB and computer architecture fields. The DB professor was so renowned that it was difficult to meet him. In contrast, the professor who became my advisor in computer architecture had not accepted students for three to four years, but the timing was right when I applied. That is how I came to major in computer architecture at Georgia Tech (Georgia Institute of Technology).

―How was life in the United States?

▲I lived in the U.S. for about six to seven years. I gave birth to my child right after arriving, so I postponed my (Ph.D.) admission a bit before starting. I was able to gain a lot of experience while I was young, not just academically. I learned that there are many different people and perspectives in the U.S., which was a great help to me.

―After returning to Korea, you joined Samsung Advanced Institute of Technology (SAIT). You must have had several options.

▲At the time, I had received several offers in Korea, but I wanted to try working at SAIT. My husband had previously worked at Samsung, and (Samsung founder) Lee Byung-chul had established SAIT in 1987 with a grand vision, and the organization was still thriving. I had the opportunity to continue my research in my major, process architecture, there.

I stayed at SAIT for about 12 years. There, I developed digital signal processors (DSPs) and transferred the technology as intellectual property (IP) to business units, leading to commercialization in mobile phones. Based on these achievements, I was able to continue my work. Later, I developed graphics processing units (GPUs). At the time, this was a significant investment for the company. Although we did not ultimately commercialize the GPU, as its importance was rising, I moved to the System LSI Division of Samsung Electronics, which focuses on system semiconductor development, aiming for commercialization.

―Back then, GPUs did not receive as much attention as they do now, correct?

▲Today, with the advancement of AI technology, the importance of GPUs in data centers has skyrocketed. However, at that time, GPUs were becoming a performance benchmark in mobile devices. Apple had started developing its own GPU, and we were also working on GPU development. There was also the question of what to do with the next-generation GPU, so at SAIT, we began developing neural processing unit (NPU) architectures, marking my first involvement with NPUs. Later, when I moved to the business division, I was responsible for GPU development for integration into Exynos (Samsung Electronics' application processor).

―After leaving Samsung Electronics, you went to Seoul National University and then to SK Telecom. What was the background?

▲While at the Department of Electrical and Computer Engineering at Seoul National University, I was working on a project related to the Samsung Neural Processing Research Center (NPRC), where I conducted NPU-related research. Later, I developed a server-type NPU as a government project, which SK Telecom led. At that time, SK Telecom was searching for the right person for their server-type NPU business. I received an offer and decided to join, not because I wanted to work at a large corporation, but because I wanted to take the lead in development.

Soojung Ryu, Visiting Professor at Seoul National University (former CEO of Saffion Korea), is being interviewed by Asia Economy at the Haedong Advanced Engineering Building, Seoul National University, Gwanak-gu, Seoul.

―As CEO of Saffion (an AI semiconductor company and former SK Telecom subsidiary), you played a major role in launching new products. Did you face any difficulties?

▲When I was only at large corporations, I did not realize it, but after leaving, I found that even small companies need to have all their operations in place, which was challenging. For example, when trying to use an enterprise resource planning (ERP) system like those in large corporations, it was too expensive. So, we had to create similar systems ourselves, and growing the company as a whole was more difficult than expected. There were also many things to consider, as we had to stick to the roadmap promised to investors.

Still, since SK Telecom was at the top of the service chain, it was good to be able to develop products while looking at actual services. For example, after developing the NPU, we were able to apply the product to a nationwide spam filtering service. On the other hand, when working at a hardware company for a long time, you tend to focus on creating the highest specs, not necessarily because they are needed, but because you have to surpass others numerically. As a result, you sometimes end up retrofitting products after they are made. This is an inherent issue with hardware companies.

―Did you continue your research on processing-in-memory (PIM)?

▲I conducted PIM research as a national project at Seoul National University and also at SAIT. While at Saffion, I collaborated with SK Hynix on PIM research. Even now, I continue to work on PIM-related projects at the university, so I remain very interested and involved in this area.

―What are your current interests?

▲Our field’s top conference will be held in October, and I have decided to organize a workshop and a competition at that time. One of my main interests now is connecting heterogeneous computing and building an ecosystem. The project I am working on is an "AI benchmarking (performance evaluation) tool." Since the scenarios for evaluating hardware can vary, I am very interested in how to approach this. Even if a product is released, proper evaluation is essential for users to trust and adopt it. Currently, there are few experts on these products, and since most are still in development, people find them difficult to use. I plan to define six industries and create benchmarking scenarios for each, so they can be utilized by industry. Several NPUs have been developed in Korea, but for activation, proper evaluation is necessary. I plan to start this way and develop tools to enhance convenience.

―Although Korea is making progress in the NPU field, its system semiconductor competitiveness remains weak. What measures are needed for growth?

▲Korea's share in the global system semiconductor market is less than 3%. While areas like AI semiconductors are growing with national attention, there are other core technologies in system semiconductors that also need continuous nurturing as foundational technologies. Additionally, Korea can be competitive in packaging and back-end process fields within system semiconductors, so these areas should be fostered. Even if these are not industry-leading fields right now, consistent policy efforts are necessary.

Not only talent development but also fostering companies is a challenge. When I returned to Korea, there were no companies developing processors, so I struggled to find where to go. Talent is not cultivated only in schools. There needs to be an industry to nurture talent, but since most are concentrated in the United States, Korean talent also tends to go there. More companies need to be established domestically to train talent, who can then start new companies and further develop the workforce. I believe it is also necessary to consolidate the various talent development projects scattered across different fields into one.

Soojung Ryu, Visiting Professor at Seoul National University (former CEO of Saffion Korea), is being interviewed by Asia Economy at the Haedong Advanced Engineering Building, Seoul National University, Gwanak-gu, Seoul.

―It must have been difficult to raise a child while continuing your career.

▲I relied on my mother for help. I commuted daily between Mokcheon, where my parents lived, and Giheung, where my company was located, until my child was in third grade. After my child entered fourth grade, we moved closer to the company, and things became more challenging. My husband and I coordinated our schedules each week to decide who would come home early on which days to take care of our child. Since our home was close to the company, I would go home after work, feed my child, send them to after-school classes, and then return to the office to work until late at night. Engineering work requires a lot of time, so it was not easy, but I enjoyed it.

―Many women struggle to balance career and childcare.

▲Family support is very important. If your child needs you, you should return home, regardless of how many children you have. Otherwise, it is okay to get help from others. It is fine to spend your entire salary on childcare. The work you do now ultimately builds your career and expertise. You are not losing anything; it all remains with you.

―What are your mid- to long-term goals?

▲I hope Korea will have at least one leading processor. I will do whatever it takes to help such a processor company exist. Also, since the AI semiconductor ecosystem is lacking, I want to explore how to connect and utilize products efficiently. Sometimes using an NPU is better, and other times PIM is more effective in a system, so I want to contribute to the expansion of AI systems from the perspective of how to best utilize them.

―Any advice for your juniors?

▲At school, I see many students whose goal is simply to get a job. Many pursue a master's degree because it is difficult to find employment with just a bachelor's. What I want to tell them is to set longer-term goals. Even if the goal seems distant and unclear, once you set one, you can check whether you are on the right path. Also, I hope they try many things. Now is the time when you have the least to lose. Opportunities lie in unexplored paths.

▶About Professor Soojung Ryu

Born in 1971, she earned her Ph.D. in Electrical and Computer Engineering from Georgia Tech in the United States. After returning to Korea, she joined Samsung Advanced Institute of Technology (now SAIT) in 2004, where she worked on DSP and GPU design. She later became an executive overseeing GPU development at the System LSI Division of Samsung Electronics. After leaving Samsung, she served as a visiting and industry professor at the Department of Electrical and Computer Engineering at Seoul National University, researching NPUs and PIM. In 2021, she joined SK Telecom, and in 2022, became CEO of Saffion Korea, which spun off from SK Telecom. She is currently a visiting professor at Seoul National University, focusing on next-generation semiconductor research.