Based on self-trained fine-tuned data analysis large models, we construct enterprise data models for different industries. Our data science team leverages deep thinking with open-source LLMs to train RAG fine-tuned models specific to enterprises, generating AI-driven data operations stewards more suitable for enterprise reinforcement learning.