Growing up, Kazuki Onodera spent his free time in video game arcades strategizing gameplay tactics. He never imagined that his passion for gaming would one day help him land a spot competing with the Kaggle Grandmasters of NVIDIA, or KGMoN.
Kaggle is the world championship for machine learning, a virtual stage for data scientists to share their knowledge and techniques through competitions, courses and discussion forums.
Onodera, now a senior deep learning scientist at NVIDIA, first discovered Kaggle in 2015 while building predictive models of Japanese banks as a financial consultant. He entered a competition recreationally to practice data modeling and found it to be a terrific outlet for creative problem-solving.
Four years and five gold medals later, he became a grandmaster — a title held by just 300 of the community’s 19 million learners.
Onodera, based in Tokyo, has been part of the KGMoN team since its founding in 2020. The nine data scientists and engineers, who hail from France to Brazil, compete in challenges to broaden access to and test the capabilities of the latest accelerating computing methods.
“When I heard NVIDIA was forming this team, I was surprised my hobby could make a real impact,” Onodera said. “Since joining, I’ve worked on models simulating climate patterns, predicting RNA molecular structures and interpreting MRI images of the brain.”
During the pandemic, Onodera built a model predicting the degradation rate of the COVID-19 mRNA vaccine. His gold-medal-winning solution, which calculated the heat sensitivity of the RNA molecules, helped ensure an effective global distribution of the vaccine.
For Amazon’s KDD Cup 2023 competition, Onodera teamed with KGMoN members across four continents to build multilingual recommender systems — winning all three tracks of the prestigious challenge. The team worked seamlessly across time zones, exchanging their global perspectives. They also swept all five tracks in the Amazon KDD Cup 2024.
“At NVIDIA, I constantly have the opportunity to build cutting-edge solutions while learning from talented colleagues around the world,” Onodera said. “I can use my passion for knowledge-sharing to advance technology.”
In addition to participating in competitions, KGMoN members publish open-source code, engage in discussions and create tutorials for developers. Their insights help guide engineering projects such as RAPIDS and NVIDIA NeMo, which help advance generative AI and data science.
“Providing valuable feedback that empowers data scientists globally has been my greatest pride,” Onodera said. “Machine learning competitions used to only be a hobby, but now, it’s become a part of my life’s work at NVIDIA.”
Learn more about NVIDIA life, culture and careers.