The objective of this project is to develop a prototype framework for deploying offline, subject-specific AI assistants, with a focus on GIS-related applications. The framework enables AI-driven GIS code generation, technical Q&A, and dual-mode interaction while ensuring data privacy through offline deployment. A key aspect of the project is evaluating the accuracy and performance of the subject-specific AI model compared to general-purpose AI models. To achieve this, a document parsing module will be integrated, allowing users to control the training material by selecting specific documents, such as GIS course materials or books, for fine-tuning the model.
Languages: Python
Suggested Software & Libraries: PyTorch, Qwen, Unsloth, Hugging Face (Model Hosting)