An interview with Robin Jeong

1. Can you introduce yourself briefly?

Hello, I am Robin, working as a deep learning algorithm engineer at Chips&Media. I majored in Math in both undergraduate and graduate school, and currently, I am working as a research agent in the company as a preferential military service.

2. What kind of company is Chips&Media?

Chips&Media is the only multimedia company in South Korea. As a side note, our company does not directly deal with semiconductors but design IPs to make semiconductors work (correspond) well. The company mainly deals with video codecs, but I have been developing hardwired deep learning IP for computer vision since I joined.

3. Why did you apply to Chips&Media?

After I decided to graduate with my master's degree, I started to study deep learning and thought it would be great to work somewhere to combine math and deep learning. At first, I self-studied by watching deep learning tutorials, thinking it's a part of just simple image processing. As I went through the sessions, I realized many applications called deep learning and got more interested in image processing.

I came across Chips&Media's recruiting announcement by chance and decided to apply for the position because it was related to image processing called super-resolution. Importantly, preferential military service was also possible.

Before the interview, I browsed the company info and noticed that the employee satisfaction rate was high. Even though I did not fully understand the business model, I thought I should check it myself. During the interview, I briefly introduced my thesis, and many of the interviewers showed interest and engagement, which I did not expect. I felt the positive atmosphere and warmth and accepted the job offer.

4. Can you tell us about your current work?

I am currently developing deep learning SR (super-resolution) algorithm in the computer vision part at Chips&Media. I think I should briefly explain what super-resolution is for a moment. I am sure everyone who reads this right now had experienced a blurry or broken picture when tried to enlarge a small image. Basically, enlarging a small resolution image to bigger clearly is called "super-resolution (SR)." When you resize the small image to a bigger size, all you have to do is drag or click. But actually, this process is only happening based on the countless experiments of so many engineers in the past. But despite the efforts, when the small size image is zoomed in, you can see the quality of an image isn't so great. Meanwhile, with the deep learning boom around the world, research on SR is also underway. There are currently studies trying to change the Heuristic function into a deep learning model, and based on the research paper, we can see that the performance of the deep learning models is much more significant.

The CV (computer vision) part that I am a part of is also developing an SR model using deep learning. Our team is developing a new model by benching marking the published deep learning SR papers. You may say why not using the best performance model from the published paper, but those kinds of models are usually quite heavy to implement, so we can't apply it right away. Therefore, we are focusing on lighting and optimizing the model. This is when Chips&Media's strengths shine. Eventually, it has to be implemented as hardware, so if algorithms and hardware are developed separately, they will not be able to optimize easily, however since we can develop it together, our company is speeding up the development process of the best model by considering these points at the same time.

5. What can you say about Chips&Media?

I really like the office atmosphere - you can ask any questions that you are not familiar with. I think this office culture is created by calling each other by the English nicknames, not by the titles. Therefore I can easily reach out to the senior engineers and interact with them comfortably.

6. When did you feel most rewarding by working at Chips&Media?

When I first came across super-resolution, I thought it was just enlarging a small image into a big picture. However, as I am working in the field, I see that there are so many different areas where SR can be implemented. Super-resolution can be applied to, not only as an image scaler, but on TVs, surveillance cameras, automotive, and much more. After knowing that it's closely related to our lives and can impact seeing images/videos, I am more actively working to develop this technology. It will be so much rewarding as our technology more widespread in society and meet the applied products in life.