Coffee Chat #11: Academic research meets product-driven work in industry

‍Q: You’ve worked on complex problems that require patience, abstraction, and long-term thinking. How do you personally stay motivated when progress feels slow or uncertain?  

A: For me, there are usually three ways to stay motivated, and I’ll give some examples from my work.

1. Discuss more. This is important because you can get new ideas simply by talking with other people. For example, when I was researching a new algorithm for packing 3D geometries efficiently, one of the breakthroughs came after discussing the problem with a friend from a completely different field during lunch. This led me to discover new tools outside my current field that I wouldn’t have known otherwise. Discussing problems with others also helps you organize your thoughts. To explain a problem to someone else, you must understand it well yourself. I found this very helpful even when the discussion didn’t yield immediate or obvious outcomes.
2. Read and learn more. Similar to my first point, the purpose of reading is to gain new ideas. Usually, when I encounter a technical roadblock, I try to read new research papers or blogs relevant to the problem. For example, during my Ph.D., I faced a technical roadblock where the algorithms I knew couldn’t produce stable or useful results. The project was stuck on this issue for almost two months. The final breakthrough came from a paper that I had previously glanced over and thought was irrelevant. Upon a more careful second reading, I found a technique in the paper that could be used to solve our roadblock, and after that, the project progressed smoothly.
3. Work on other projects. When you feel stuck and are making no progress, switching to another project can sometimes be a good choice. For example, when I joined Inkbit, I used an existing and well-known algorithm to solve a packing problem. The results were not optimal, but they were good enough for practical use. We wanted to improve it further, but after making no progress for a month, I ran out of fresh ideas. After discussing this with my superiors, we decided to switch to a new project. I didn’t return to the packing project for quite some time, until I came up with a new idea after talking about it with friends. So, it is okay to switch to other projects and come back later when you find new ideas!

 

Q: You’ve contributed to both academic research (e.g., SIGGRAPH publications) and product development at Inkbit. How do you balance research creativity with product requirements?

A: There is a complicated interplay between research creativity and product requirements. Usually, product requirements come first.

For example, one of our clients wanted to use our algorithm for a different application. The client provided very detailed requirements and data for evaluation. I conducted an early assessment and found that our algorithm produced results that were about 10% worse than the required target, meaning a non-trivial extension was needed to bridge the gap.

As a result, we needed to carefully decide whether to invest in this project. First, we discussed with the client to estimate the potential revenue from the project. We also conducted market research and found that it could be commercially feasible. Then, I started the technical evaluation. I needed to determine what techniques could be used to bridge the performance gap and how much time would be required to implement these changes. I spent two weeks researching and reading relevant papers. Fortunately, I found a decisive paper that gave me confidence that I could improve our algorithm to meet the client’s requirements. After that, the timeline became clear.

We ultimately decided to move forward with the project. This early research planning was crucial to meeting the client’s timeline. After five months of work, the client was very happy with the results.

Q: In your experience teaching computer graphics (WebGL/Three.js) and mentoring students, what do you find are the most meaningful skills or mindsets that differentiate successful learners in this field? ‍

A: There are a few things that really make a remarkable difference. In computer graphics, everything is usually connected to visuals in some way. Therefore, students who are capable of quickly visualizing data tend to produce better projects. I taught some of these visualization tricks; for example, mapping data to the color buffer in a shader, where the data can be visualized as color on the screen. This is much better than printing data to a text output or console and then looking at the values, which is commonly how it’s done in many other programming contexts. In computer graphics, however, data is usually in the form of images, colors, or 3D shapes. Being able to visualize this data quickly and develop an intuition makes a big difference in this field.

Another important mindset is the ability to progressively build complex projects from simple steps and validate each step. In one of my assignments, I asked students to build a complicated shader. I demonstrated how to start from a very simple example, such as drawing a basic triangle. Students then progressively built more complex geometry and finally added animations. Because shaders compile and run on GPUs, they are much harder to debug. I asked students to extend this shader to draw more complex shapes or animations. I found that students who followed these atomic steps were more likely to complete the assignments successfully with less help.

I found that these mindsets are also very helpful in my daily work. For example, being able to visualize complex data helps me summarize information and draw conclusions. Being able to decompose a complicated project into small steps helps me stay motivated and on schedule throughout the project. I hope that students who attended my project will be able to apply these skills to their studies and jobs as well.

‍Q: Looking ahead, what emerging challenges or opportunities in computational graphics, simulation, or optimization are you most excited about, and how do you think the field will evolve over the next 5–10 years?

A: There are a few interesting opportunities in the future of computer graphics and visualization that I’m excited about, but I’ll focus on one direction.

This direction is the application of these technologies to a wider range of industries. Over the past few decades, researchers have developed sophisticated computer graphics and visualization algorithms, primarily for the film industry (e.g., Pixar and Disney) and the video game industry. One emerging trend is to leverage these techniques for industrial digitalization: digital twins. Nvidia has been pushing hard in this direction. For example, they have collaborated with Amazon to build digital warehouses that simulate entire warehouse operations. This allows companies to accurately simulate their warehouses, closely monitor performance, and optimize operations.

I believe a large segment of the manufacturing industry can benefit from digital twin simulations, where accurate graphics and physical simulation play pivotal roles. At Inkbit, one of the products we are developing is closely related to this trend. We collaborate with packaging engineers to develop a tool called “Pack-Studio,” which creates digital package designs for the automotive industry. Packaging engineering is a long-established field that still relies heavily on legacy tools. We found that our clients greatly appreciate the improved visualizations provided by this new tool.

‍Q: What is some advice you have for students who want to pursue a Ph.D. degree?

A: The first piece of advice is to figure out why you want to pursue a Ph.D. and to check whether you are financially prepared for it. Doing a Ph.D. is a very daunting task. You are likely to spend five or more of your best years in a lab, conducting experiments, reading and writing papers, and occasionally presenting your work (usually one of the best parts of a Ph.D.). You will also be paid far less than you would be in a full-time job. You need to genuinely enjoy open-ended problems and have a strong desire to push the boundaries of human knowledge.

I chose to pursue a Ph.D. because I truly enjoy tackling hard problems. The “aha” moments and presenting research to hundreds of people at conferences are among the best experiences of my life. If you are still an undergraduate, one way to test whether you have the passion and patience for a Ph.D. is to take graduate-level courses. Many of these courses include a project component, where you dive deeply into research papers and implement the details. If you find yourself enjoying the process, you may be well suited for a Ph.D.

If you still decide to pursue a Ph.D. career, the most important factor in your research is your advisor. You should look for a supportive and “well-funded” advisor (very important). This is not always easy; one way to learn more is to reach out to a potential advisor’s current students and gather their insights. You will also need to plan to get strong recommendation letters. Trying to join a research group after completing most of your coursework can be immensely helpful.

‍Q: What are some differences in conducting research in academic vs. industrial environments?

A: I’ve conducted research in both academic (e.g., during my Ph.D.) and industrial (at my current company) environments, so I can discuss the differences across several aspects.

In an academic setting, the most important purpose of research is publication. Practical applications may be considered, but they are usually not the primary factor. For example, during my Ph.D., I researched a fluid simulation technique that was quite niche and did not have wide industrial applications. Nevertheless, the results were interesting, and we were able to publish two top-tier papers on the topic, which satisfied my Ph.D. graduation requirements :).

After I joined my current company, I was tasked with researching a new method to efficiently pack 3D geometries. The purpose of this project was purely practical, focusing on its application in 3D printing rather than on publication. After obtaining strong results on the packing project, we later decided to publish the work to attract more attention to the company, which is like icing on the cake. This happened long after the technique was already in practical use in the company and patented.

In summary, in academia, the most desirable outcome of research is typically publication in well-known journals or conferences. In industry, the most desirable outcome of research is practical application in the field.