Advanced machine learning (ML) models have been developed for battery lifetime prediction in different use cases at all stages of a battery's life. As the first step to enable the transferability of ML models for battery lifetime prediction across multiple use cases, a scenario-aware machine learning pipeline is proposed, in which two feature engineering methods that have been able to generate input features with outstanding predictive power are used to learn the best ML model for battery lifetime prediction in a chosen usage scenario. The experimental results show that the histogram-based feature engineering method is able to generate input features with predictive power generalized across two usage scenarios (i.e., identical cycling and protocol cycling). Thus, to enable transferability of ML models for battery lifetime prediction across different scenarios, and even battery chemistries, this histogram-based feature engineering method will be further investigated together with online fine-tuning strategies.