Tea leaf diseases pose a significant threat to the economically vital tea industry in Assam, India. While deep learning models have shown high accuracy in academic settings for disease detection, a persistent gap remains in translating these models into accessible, field-deployable tools for farmers. This paper addresses this research-to-practice gap by presenting a methodology for the cloud-based web deployment of a validated, high-performance hierarchical two-stage deep learning framework. The model, trained on region-specific data from Dibrugarh, Assam, achieves 96% accuracy in classifying key tea leaf diseases. The proposed deployment architecture leverages Hugging Face Spaces for automated containerization and hosting, paired with a Streamlit-based web application for an intuitive interface. This lightweight strategy simplifies deployment, making the tool readily accessible to non-technical stakeholders via a standard browser. The result is a reproducible, scalable, and user-friendly solution that bridges the divide between algorithmic innovation and agricultural application, offering a tangible decision support tool for tea farmers.