Finetuning framework for Pillar medical imaging models.

This repository contains the pretraining code for the Pillar-0 model.

A comprehensive evaluation pipeline for Vision-Language Models on medical imaging tasks, with built-in support for multi-GPU processing, real-time progress tracking, and disease finding classification.

A Python-based engine for processing radiology reports using the Qwen3 model with sglang for efficient batch inference. Includes quality control file generation and performance evaluation tools for comprehensive validation workflows with debug mode for faster iteration.

A high-performance medical image processing engine for converting DICOM and NIfTI files into ML-ready formats with intelligent windowing and compression.

Human-Level Information Extraction from Clinical Reports with Fine-Tuned Language Models