It typically includes model selection, parameter tuning and optimization, use of pre-trained models and transfer learning, right use of limited data, model interpretability and explainability, feature engineering and autoML robustness and security, and computational cost â€" efficiency and scalability.