Mozu Research
The full context of care, not isolated text.
Mozu builds healthcare datasets for training and evaluating frontier AI models.
Our research datasets are collected from real clinical care in Japan and combine patient interactions with the medical records, clinical decisions, treatments, and outcomes that follow. This gives models access to the full context of care rather than isolated medical text.
Research areas
- Clinical reasoning
- Medical language
- Multimodal understanding
- Longitudinal care
- Healthcare task execution
Datasets
Doctor Patient Conversations
100,000 telehealth conversation transcripts linked to the doctor’s notes, medical records, and outcomes.
100,000
Telehealth transcripts
Dataset overviewJapanese Medical Audio
Japanese medical audio with transcripts, speaker diarization, and review by doctors in Japan.
12,500
Hours of Japanese medical audio
Dataset overview
We work with AI research teams on model training, evaluation, and reinforcement learning. Full dataset pages are restricted and require sign-in.