Data Scientist from Matsuo Institute Wins Gold Medal in Kaggle Competition, Ascends to "Kaggle Competition Grandmaster"

Yuki Okumura, a data scientist at Matsuo Institute, won a gold medal in a Kaggle competition for translating Akkadian, earning the elite Grandmaster title held by only 0.2% of competitors.

📋 Article Processing Timeline

  • 📰 Published: April 1, 2026 at 23:00
  • 🔍 Collected: April 1, 2026 at 16:47
  • 🤖 AI Analyzed: April 17, 2026 at 03:10 (370h 22m after Collected)
Yuki Okumura, a data scientist at Matsuo Institute Inc. (hereinafter "Matsuo Institute"), has won a gold medal in the "Deep Past Challenge – Translate Akkadian to English" hosted by Kaggle, the world's largest AI competition platform. With this competition, Okumura has been certified as the 393rd Competitions Grandmaster in the world. This corresponds to the top 0.2% of the approximately 200,000 active competitors.

"Kaggle" is an online platform for competing in challenges related to data analysis and machine learning. Data scientists and statisticians from all over the world participate, which is instrumental in developing AI talent and solving corporate problems. Many of the world's top-level data scientists and machine learning engineers participate in this competition, and we believe that this award is a global recognition of the creativity and technical capabilities of Matsuo Institute's data scientists.

## Initiative Details

In this competition, participants were tasked with targeting cuneiform documents left by ancient Assyrian merchants about 4,000 years ago, and building a model to translate transliterated Old Assyrian (a dialect of Akkadian) into English. Participants competed on how accurately they could translate commercial records and letters into English by designing and implementing a translation model capable of handling an ancient language with highly limited data and complex grammar.

## Award Winner's Comment

**Yuki Okumura, Data Scientist, Matsuo Institute Inc.**

Because Akkadian is a low-resource language, expanding the training data was absolutely essential. Therefore, we carefully extracted a parallel corpus while organizing unstructured data in PDF format using Vision-Language Models (VLMs). Furthermore, we utilized methods that help improve the performance of translation models, such as continuous pre-training and the addition of back-translation data, which led to an overall improvement in accuracy. I hope that this competition will act as a catalyst to further advance the analysis of the Akkadian language.

Reference Article: "Able to solve problems in various business domains" - Why a Kaggle Master joined Matsuo Institute.

## Regarding Recruitment

At Matsuo Institute, top-level data scientists, including Kaggle Grandmasters, are enrolled and work on the social implementation of AI while continuously challenging and improving themselves every day. We would like to meet people who share our aspirations and want to challenge themselves at an even higher level.
Please feel free to apply from the job openings below. Let's start with a casual interview to get to know each other.

▼ Job Position
Senior Data Scientist: https://open.talentio.com/r/1/c/matsuo-institute/pages/77463

## About Matsuo Institute Inc.

Matsuo Institute Inc. is a research institute established with the aim of working alongside the Matsuo-Iwasawa Laboratory, Graduate School of Engineering, The University of Tokyo, to create and greatly develop an "ecosystem" that generates innovation centered around the university. Matsuo Institute carries out the "development and implementation" of research results and technologies generated from academia, aiming for their widespread dissemination in society and contributing to the improvement of Japan's industrial competitiveness.

About Matsuo Institute: http://matsuo-institute.com
About us

Keywords: AI, Kaggle, Grandmaster, Data Scientist

FAQ

What is a Kaggle Competition Grandmaster?

It is the highest-ranking title awarded on Kaggle, achieved by only the top 0.2% (roughly 400 individuals globally) of its 200,000 competitors.

What was the task for this Kaggle competition?

The objective was to build a machine learning model capable of translating 4,000-year-old cuneiform texts (Old Assyrian/Akkadian) into English.

What kind of organization is Matsuo Institute?

It is an organization aligned with the University of Tokyo's Matsuo-Iwasawa Lab, dedicated to the development and social implementation of advanced AI technologies.