This repository provides an end-to-end training data generation and synthesis framework for domain-specific AI reasoning, centered on a deep research agent for medical scenarios. The system consists of three integrated components: a knowledge graph construction module that transforms medical domain knowledge into high-quality QA pairs with automated reasoning paths; a trajectory generation pipeline that converts these pairs into multi-turn reasoning trajectories with tool interactions; and an evaluation framework for assessing model performance. The framework produces training data for a 32B parameter medical reasoning model.