Machine learning (ML) is part of artificial intelligence (AI) that applies ML models to train and process data from multiple sets of datasets. However, training ML in-browser is still problematic, particularly in areas dealing with sensitive user data. This research investigates the feasibility and limitations of training ML models directly within web browser environments, using advances like the WebAssembly (WASM), WebGPU, and JavaScript ML libraries. The study examines the impact of dataset size, model complexity, and resource constraints on training performance and efficiency, while also considering the privacy advantages of client-side computation. The findings of this study provide valuable insight into the current limitations of browser-based ML and provide information to guide the future development of privacy-aware and performant web applications.