I am a PhD Student at Institute of Science Tokyo (formerly Tokyo Tech). My research lies at the intersection of HPC and Machine Learning, specifically focusing on distributed training and low-precision training (FP8/NVFP4) for Large Language Models.
I am a core contributor to the Swallow Project, a Japanese LLM development initiative, where I maintain the pre-training library and lead large-scale training experiments.
- [Jan 2026] My paper "Rewriting Pre-Training Data Boosts LLM Performance in Math and Code" has been accepted to ICLR 2026! 🎉
I am actively looking for Research Internship opportunities in the US If you are interested in my work on LLM systems and low-precision training, please reach out!
- 🌐 Website: https://okoge-kaz.github.io/
- 🎓 Google Scholar: Citations Profile
- 💼 LinkedIn: kazuki-fujii
- 🐦 X (Twitter): @okoge_kaz