Japanese indirect prompt-injection datasets (translation-first). Hub: github.com/masahirocom/japanese-llm-security
masahiro nakatsugawa PRO
masahiroid
AI & ML interests
AI architect and researcher focusing on Japanese language models, embeddings, RAG, and applied AI systems.
My work spans Japanese semantic analysis using Ruri, domain-specific LLMs, synthetic dataset generation, AI-driven marketing intelligence, multilingual news analysis, and cultural knowledge systems.
I am particularly interested in embedding model evaluation, fine-tuning, semantic clustering, knowledge representation, and the integration of LLMs into production applications.
I also work on AI-assisted software engineering, evaluation frameworks, and AI governance, with a focus on building reliable and practical AI systems.
Recent Activity
liked a model 4 days ago
masahiroid/Qwen3.8-27B-Pyrite-2bit-MLX updated a model 4 days ago
masahiroid/ruri-v3-70m-mlx published a model 4 days ago
masahiroid/ruri-v3-70m-mlx