v0.1.0

First public release of MOOZY, a patient-first foundation model for computational pathology. This release includes slide and case-level encoding via CLI and Python API, the full two-stage training pipeline (self-supervised slide encoder + patient-aware multi-task alignment) with local scripts and SLURM job templates, and a pretrained checkpoint that downloads automatically from HuggingFace on first use with no access gates. 85.77M parameters, 14x smaller than GigaPath, trained on 77,134 public slides across 333 tasks from 56 datasets.

• Project Page
• Paper
• HuggingFace Model
• PyPI
• Documentation: Encoding
• Documentation: Stage 1 Training
• Documentation: Stage 2 Training
• Model Card