Recent Updates(12 items)
A side-project paper I co-authored, "Structuring Cross-Domain Inquiries for Small-Scale Live Events through Inductive Foresight and Impact Dynamics," was accepted to the 2026 IEEE International Conference on Industrial Engineering and Engineering Management (IEEM 2026), 2–5 December 2026 at Marina Bay Sands, Singapore.
A side-project paper I co-authored, "Gender Bias in LLM Hiring Decisions: Evidence from a Japanese Context and Evaluation of Mitigation Strategies," was accepted to the BIAS 2026 Workshop at ECML PKDD (September 7, 2026).
Lead Researcher – Shiba AI (Shibaura Institute of Technology)
Joined Shiba AI, a rapidly growing Japan-based AI safety lab at Shibaura Institute of Technology, as Research Project Lead. Focused on Mechanistic Interpretability — understanding the internal mechanisms of large language models through circuit-level analysis and interpretability methods.
Experience
2026/4–Present
Current- Joined the Research and Development Center for Large Language Models (LLMC) in April 2026
- Research focused on Post-training
- Official appointment as Technical Assistant beginning June 2026
2025/5–Present
Current- Co-Researcher at Matsuo–Iwasawa Lab
- Focusing on Mechanistic Interpretability, Post-training, and more
2026/2–Present
Current- Leading research and development of Physical AI for a new business initiative
- Driving technical direction from the ground up, bridging cutting-edge AI and Robotics
2025/8–Present
Current- The world's largest university-based consultancy for non-profits and social enterprises
- Delivering strategic and operational consulting to socially conscious organizations
Lead Researcher – Shiba AI (Shibaura Institute of Technology)
2026/4–Present
Current- Research project lead at Shiba AI, a rapidly growing Japan-based AI safety lab focused on Mechanistic Interpretability
- Understanding the internal mechanisms of large language models through circuit-level analysis and interpretability methods
2025/11–2026/1
- Taught LLM Analysis & Interpretability to graduate students and industry professionals
- Covering cutting-edge technologies essential for implementing LLMs in societal and industry contexts
2025/5–2026/1
- Founding CTO at a smart parking startup transforming parking infrastructure in Japan
- Led end-to-end technical development from zero to first product release
- Overseeing full-stack architecture, deployment, and product strategy
2025/5–2026/2
- Developing application and system for better franchise management of restaurants
- Leading development teams and coordinating cross-functional collaboration
Research
I work on the training and interpretability of large language models — how they acquire capability after pre-training, and what actually happens inside them.
Results from several ongoing projects are forthcoming.
Papers
Haruki Yajima, Rihito Kotani*, Nozomi Fukawa*, Nonoka Suzuki, Mizuki Sekiuchi, Kyosuke Hino, Hajime Sasaki, Yukihiro Nakajima
* equal contribution
Accepted at IEEE IEEM 2026December 2026 · Singapore
Serena A. Hoffstedde, Machiko Hirota, Akshara Nadayanur Sathis Kanna, Rihito Kotani, Ujwal Kumar, Gabriele Trovato, Phan Xuan Tan
Accepted at BIAS 2026 Workshop @ ECML PKDDSeptember 2026arXiv:2606.18649
Education
April 2025 – March 2029
CurrentMember of the Nakazawa–Ōkoshi Research Laboratory
September 2023 – December 2023
Stanford e-Japan Program (SPICE), Fall 2023. Honorable Mention recipient.
Tokyo Gakugei University International Secondary School
Graduated March 2025
British School of Yangon, Myanmar
2017 – 2021
Attended from 2017 to 2021 (Primary 5 through end of Secondary)
Awards & Honors
Highest award from Japan's Ministry of Education for outstanding student research achievements
World's largest pre-college science competition. Received special award in Computer Science category for LLM interpretability research
One of Japan's most prestigious student science competitions, awarded by the Minister of Science and Technology Policy
National business plan competition organized by Japan Finance Corporation. Placed 3rd out of thousands of entries
United Nations Development Programme initiative recognizing young social entrepreneurs in the Asia-Pacific region
Stanford University's program for outstanding Japanese high school students studying U.S.–Japan relations
National app development competition for high school students in Japan
Selective program on AI and human interaction
Selected participant at Asia's largest student-run conference
Skills
- AI / ML: Large Language Models, Interpretability, Physical AI, Applied Machine Learning, Model Training & Fine-tuning
- Development: Full-Stack (Python, Go, JavaScript/TypeScript, SwiftUI, mid level cpp)
- Cloud & Infrastructure: AWS
- Other Strengths: Start-up product development, consulting, research communication, agile teamwork
Additional Information
- Bilingual: Fluent in Japanese and English
- Lived in Myanmar (2017–2021); experienced in international education and diverse collaboration
- Strong experience in public speaking, pitch competitions, and leadership roles in both research and product development