“Only one debate per day?
Not enough! Prepare more whiteboards!”
- An interview with Dr. Seulki Yeom, project manager of Nota's research team.
The word research often has implications of passivity. We imagine someone sitting alone at a desk, quietly focused with a furrowed brow, reading and writing for hours on end. If we were to assign an MBTI test result to a typical researcher, we would most likely guess that they were an I, introverted, and not an E, extroverted. Nota's research team subverts these expectations. In a field where telecommuting is on the rise, Nota's research team continues to go to the office, perhaps even more than anyone else in the company. They work on projects together, discussing and debating ideas, with the shared goal of progressing Nota's vision for the future. Nota's research team is based in Berlin, Germany, within the European subsidiary of the company. Just like Berlin, the research team is modern, dynamic and forward-thinking.
The office is a flurry of activity. Conversation flows, debate ensues and there is rarely a moment's rest. Whiteboards are scrawled upon again and again throughout the day, like a scene from Good Will Hunting.
I met with Dr. Seulki Yeom, the project manager of Nota's research team, to discuss her work.
You once said that “research teams always fight”, could you elaborate?
It sounds worse than it is! The truth is that we don't fight with our fists, we simply have healthy debates. In a sense, it is our job to argue. “I'm right, you're wrong, and here's why” is the best way to make progress. When we publish a new thesis, or issue a new patent, critiques are inevitable. A culture of healthy debate, or if you like, fighting, keeps us on our toes.
The alternative is stagnation. If we become too comfortable or afraid to challenge one another, our work will suffer. The team's performance depends on how open and honest we are as a group. The battles are never personal, and we all have the same goal in mind. Working through problems together logically matters, and it helps us in the depth and scope of our research.
That's an interesting perspective. What topic are you fighting over right now?
Lately, we have been studying a method known as pruning. Simply put, it is a form of model compression that reduces weight. Pruning has the best weight reduction performance compared with other methods that are available, it has a lot of potential. The results were so encouraging that we wrote a thesis on the topic, which was accepted.
More recently, I have been researching the field of Neural Architecture Search, or NAS. It is a fairly challenging prospect. NAS is the most inaccessible methodology available within lightweight modeling. Most lightweight modeling research can typically be done with a single computer in a lab. If your GPU is sufficient, you can run tests and get immediate results, editing and reacting in real time. But for NAS, it takes months to design and test. It is hard work, but I am enjoying it.
It sounds a lot more dynamic and challenging than I imagined. Could you tell me a bit more about the role of the research team at Nota?
We are structured in a very particular way. On one side of Nota, you have the business development team, the strategy team, and marketing and operations. While development teams like HAI, ITS, and NetsPresso are on the other side. These teams focus on creating real products, as quickly and efficiently as possible, with the aim of bringing them to market.
Our team is different. You could think of the research team as being in the backroom, working as stewards of the forest. Our primary mission is to carry out ongoing research and solve any problems that arise. Of course, we also need to remain in constant communication with the other teams, ensuring our technology is running smoothly. It is not a flashy role that earns money in the short term, but one that tends to the garden and ensures a prosperous future.
Would you say that you are a trendsetter in the world of lightweight research?
That's not for me to say. What I can say is that we are obsessed with our work, and we will continue to push the envelope as much as possible. When the development team puts our ideas into practice, it makes it all worthwhile.
I am proud of the work that we do. When we write new papers, we always welcome the feedback of the development team. They are concerned with the prospective performance of the technology, whether it is good or bad, whether the research can be applied to a product or not. Their input gives us valuable insight from a different perspective. It is a team effort.
What would you say makes Nota's research team so special? What makes you stand out from the competition?
We commit to doing the work properly. We truly do research.
What do you mean?
Generally speaking, it is rare for a Korean company to invest in a research team. This is true across the board, from large companies to smaller startups. Independent research is expensive, and it takes time, there are no immediate results. It requires patience, and long-term commitment.
Many research papers you read today are built on the premise of “can this make money?” This is because they are primarily motivated by profit, research is just a means to that end. True, comprehensive research is often neglected. Many companies will hire one or two people for research, but that is as far as it will go. In an environment like this, proper research is impossible.
How did the Nota research team come into being?
I think this is where we have an advantage. Myungsu, our CEO, has experience working in research and understands the value that it brings to a project. This is what sets our company apart. Companies founded with a business mindset often relegate research to an afterthought. Technologies that emerge from research take time, sometimes several years. Myungsu understands this, and he saw to it that a real research team was put in place.
We are very lucky to have a CEO that understands what research entails. Research doesn't generate profits in the short term, and more short-sighted people are often turned off by this. However, with patience, research plays an important role in increasing the value of company technology over time. This can be amplified even more through patents and publications.
We understand that things take time. Rome wasn't built in a day, and a forest doesn't grow overnight. With hard work, patience and diligence, we are slowly growing our own forest.
Is there anything you are particularly proud of that you would like to share?
Over the past two years, our research team has successfully published five new papers. This has led to various opportunities, and added positive value to the company's name and reputation in the field. Furthermore, we filed two new patent applications. I am very proud of the work we are doing.
For those interested in joining Nota's research team, what qualities should they possess?
We are interested in results. Our research team is small, and we cultivate a tight-knit team that is disciplined and works hard. The ideal candidate would fit this profile.
You should have some experience of writing a dissertation. The process of writing a thesis is long and arduous. The first hurdle you have to overcome is selecting the subject, which is not as easy as it sounds. After choosing your subject, you have to research the topic and write. At the end of it all, your work must be reviewed and accepted. It is a tall order, and not many people have crossed that particular mountain. Additionally, it would be ideal if you have the ability to speak English well. English is the main company language, and it is what we use to debate! We look forward to hearing from you. Thank you.