This paper explores the potential of Tiny Machine Learning (TinyML) for privacy-preserving building energy management systems on mobile devices. While TinyML offers reduced latency and improved privacy, its effectiveness in predicting building energy consumption on mobile devices is not well studied. The proposed approach prioritizes user privacy by processing and storing energy data locally on users' mobile devices, leveraging smartphone, tablets, edge nodes, and secure cloud storage. This empowers users with control over their data and adheres to privacy regulations. Predicting building energy usage on mobile devices is crucial because it offers portability, accessibility, and privacy, as well as fosters user engagement. Mobile predictions allow users to conveniently monitor and regulate energy consumption, improving accessibility. Additionally, processing data locally ensures privacy by keeping sensitive information under user control. The paper also investigates the feasibility of converting a TensorFlow-based long short-term memory (LSTM) neural network model for energy prediction to a CoreML or TensorFlow Lite model for deployment on mobile devices. The results indicate a significant degradation in model accuracy after conversion to a CoreML and almost no degradation after conversion to a TensorFlow Lite model. Further research is recommended to explore optimization techniques for the conversion process and to compare models with other criteria.