Ricoh's Initiatives Beyond His Specialty: How a Ricoh Engineer Reached Kaggle Gold

As artificial intelligence (AI) adoption continues to accelerate, an organization’s competitiveness increasingly depends on people who can apply digital technologies to solve real-world challenges. In Japan, Ricoh supports employees in developing their digital skills through initiatives such as the Ricoh Digital Academy, which provides digital technology training and hands-on learning opportunities, as well as its professional certification programs.

Daisuke Maeda is one such employee. Originally an engineer who worked in robot mechanical design, he began studying AI and mathematical optimization motivated by the challenges he encountered in his day-to-day work. He eventually went on to earn a Gold Medal in Kaggle, one of the world’s leading AI competition platforms. We spoke with Maeda about the challenges that led him beyond his original field of expertise and the problem-solving skills that matter in the age of AI.

Estimated reading time: 8 minutes

Maeda’s journey to Kaggle began with his work in mechanical design. While developing a high-speed laser control system, he incorporated combinatorial optimization techniques into a drawing algorithm. That experience became his introduction to mathematical optimization.

“We adopted combinatorial optimization techniques for controlling a laser drawing algorithm. Through guidance from colleagues who specialized in mathematical optimization and my own study, I became fascinated by the field. I wanted to put what I had learned into practice, which is why I started competing on Kaggle.
Looking back, I think the foundation of this achievement was a series of small steps beyond my original field of expertise. Each time I faced a challenge in my work, I learned what was necessary to solve it and gradually expanded my expertise.”

Maeda participated in the Santa 2025 - Christmas Tree Packing Challenge, held from November 2025 to January 2026. The competition required a strong mathematical approach and attracted 3,357 teams from around the world. His team ultimately placed 10th overall.

The Santa 2025 challenge seemed simple at first glance: participants had to arrange Christmas tree shapes of identical size and form within the smallest possible square. A total of 200 problems were presented, ranging from one to 200 trees, and participants were scored based on how efficiently they packed the trees.

“It looked simple at first, but it was actually an extremely difficult problem,” says Maeda.

Because each of the 200 problems had different conditions, participants had to choose the most effective strategy for each one. For smaller numbers of trees, success depended on rapidly exploring different positions and rotations. For larger numbers, they first had to identify an area-efficient base pattern, then carefully determine how to place the remaining trees.

Maeda initially entered the competition on his own, handling everything from algorithm design and implementation to evaluation. However, the number of possible arrangements was enormous, making it impractical to explore every option manually. That was where AI became an indispensable partner.

“Packing problems like this have long been studied in industrial fields, such as optimizing how materials are cut. I used AI to research existing studies and related techniques, then rapidly created implementation prototypes using a vibe coding^*1 approach. But I never accepted AI’s suggestions at face value. I tested each method, evaluated its effectiveness, and selected only those that worked well for this challenge. By using AI to accelerate both research and implementation while repeatedly testing and refining my ideas, I was able to catch up on new techniques and put them into practice much faster than I could have on my own.”

Because the competition included 200 distinct problems, no single strategy could maximize the overall score.
To pursue a higher ranking, Maeda joined a four-person team that included a Kaggle Grandmaster^*2 and other highly experienced competitors.

The team shared their best-performing algorithms and high-scoring layouts while comparing which approaches worked best for different types of problems.

“Each member had a different style,” Maeda recalled. “Some excelled at running large-scale parallel computations to explore huge numbers of possibilities. Others specialized in efficiently narrowing down promising candidates. Still others focused on discovering structural arrangements for cases involving large numbers of trees. It was incredibly stimulating.”

Maeda contributed his own observations and improvements while learning from the methods and high-scoring configurations produced by his teammates.

“I was able to build on my own strengths while incorporating the strengths of others. By combining multiple approaches, we were able to reach levels that no single solution could achieve on its own.”

One reason Maeda decided to join the team was his desire to understand how top competitors approached problems.

“Their ability to identify what would truly improve the score, and what wouldn’t, was remarkable. They could quickly recognize where effort would have the greatest impact and focus on those areas.
I was also impressed by how quickly they built tools for combining different solutions and by their mindset of steadily pursuing improvements one step at a time.
They didn’t rely solely on flashes of inspiration. Instead, they accumulated many careful refinements. I learned a great deal from that approach.”

For Maeda, the value of the competition went far beyond earning a Gold Medal.

“I developed the ability to structure problems, formulate hypotheses, and repeatedly test them while searching for better solutions under constraints.
Working with people who had different strengths and methodologies taught me lessons that are directly applicable in business, where solving challenges often requires collaboration across departments and professional disciplines.”

Today, Maeda works in the MIOI (Market In Open Innovation) Promotion Group, where he works to increase the chances of turning research outcomes into viable businesses. As part of that work, he is helping develop an AI system that supports idea generation for connecting research themes to business opportunities.
Many of the lessons he learned through Kaggle now inform that work.

“During the competition, I relied heavily on AI for both research and implementation. What I realized was that even when AI can propose many possible approaches, human judgment remains essential for selecting the right one and improving it.
We should let AI handle what it does best, while humans make the final decisions. That division of roles influences how I design AI systems today.”

Maeda also believes AI can broaden opportunities for people regardless of their programming background.
“With AI, even people who have never developed software before can work more efficiently and creatively.
I would first like to create an environment where more people at Ricoh can make effective use of AI, helping to move research themes towards commercialization and support the creation of new value.”

Through Kaggle, Maeda discovered the excitement of problem solving itself: organizing conditions, developing hypotheses, conducting repeated experiments, and gradually moving closer to better solutions.

“Even in the AI era, the ability to structure problems and solve them collaboratively remains essential. Kaggle reminded me of that.”

Competing successfully alongside world-class participants also gave him confidence.

“I used to see Kaggle Grandmasters and Gold Medal winners as people operating at a level I could only aspire to,” he recalled with a smile.
“But I learned that what sets them apart isn’t necessarily extraordinary inspiration or knowledge. It’s their attention to detail and refusal to compromise.
If you continue making steady efforts, you can work alongside the world’s best. That experience gave me tremendous confidence.”

Today, Maeda is also involved in software development through an internal side-job program. Having studied robotics at university and worked across mechanical, electrical, and software disciplines, he has never been afraid to explore new fields.

“AI makes it easier to learn new technologies efficiently and take on areas where you have little prior experience. I hope to continue solving challenges both internally and externally without being limited by boundaries.”

Maeda’s journey challenges the notion that expertise must stay within fixed boundaries. By continuing to learn, embracing new fields with AI, and collaborating with others to solve problems, Maeda is finding his own sense of “Fulfillment through Work.”