Design and Development of the AI Assistant of the Drug Stores of Hamadan University of Medical Sciences
Code: G-1062
Authors: Javad Keshvari Kamran * ℗
Schedule: Not Scheduled!
Tag: Intelligent Virtual Assistant
Download: Download Poster
Abstract:
Abstract
Background and aims: The innovation of this research is the use of modern AI tools and LLMs and the absence of entering the field of neural network model design in order to increase the efficiency and usability of the system. Method: The steps of the work method are: 1) Studying suitable LLM AI models and finally selecting and using the LLAMA 3.1 70b model of Meta. 2) Preparing and completing project content files and documents such as SQL reports, Excel QA files, PDF files, etc. 3) Performing preparation and cleaning operations on the above-mentioned reports, files and documents. 4) Performing the Python development and programming process for the project such as including the libraries used are: langchain, llama_index, docker, ollama. 5) Implementing the necessary hardware infrastructure and servers with appropriate GPU. 6) Performing technical and content testing stages. 7) Final implementation. Results: Due to data security and confidentiality requirements, all components of the final application were installed and implemented on the local server. Through the ollama Python library solution, the selected AI model was downloaded to the university's local server, and then the programming process was carried out exclusively without the need to access online models. Conclusion: According to the experience gained from this research, it seems that many areas in organizations will become highly effective and productive if they use AI assistants.
Keywords
Chatbot, LLM, Drug Stores, Local Server
