Bridging Data And Hardware Gap For Efficient Machine Learning Model Scaling