http://repository.president.ac.id/xmlui/handle/123456789/13 Information Technologies 2026-04-07T14:49:47Z http://repository.president.ac.id/xmlui/handle/123456789/13856 IMPLEMENTATION OF RAG SYSTEM FOR ENHANCING THE EFFICIENCY OF HUMAN CAPITAL DIVISION IN ANSWERING EMPLOYEE QUESTIONS AT XL AXIATA P, Natalia Desy Anggreani Human Capital (HC) departments are often burdened with repetitive questions from employees regarding policies, benefits, procedures, and other administrative matters, so these routine tasks can divert valuable time and resources from strategic HC functions. This study aims to develop and implement an intelligent system that is capable of delivering accurate, relevant and context-aware responses through domain-specific chatbot using Retrieval-Augmented Generation (RAG) architecture, specifically designed to assist employees in the company policies domain. The system integrates a large language model developed by OpenAI with a curated internal knowledge base consisting of company policies and Human Capital-related documents. The RAG framework was selected for its ability to combine information retrieval with generative capabilities, enabling dynamic responses based on factual content. The findings show that the proposed chatbot significantly increased the accuracy and relevance of responses compared to traditional FAQ systems and general purpose chatbots. The domain-specific RAG chatbot demonstrates strong potential to enhance the operational efficiency of HC divisions. By leveraging OpenAI’s advanced language model, the system represents a significant step forward in intelligent employee support and paving the way for more intelligent and responsive Human Capital services. 2025-01-01T00:00:00Z http://repository.president.ac.id/xmlui/handle/123456789/13312 IMPLEMENTING DEMAND FORECASTING SYSTEM TO ENHANCE INVENTORY MANAGEMENT EFFICIENCY FOR SMES Ginting, Enjel Rupiani Br We would like to introduce a web-based warehouse management application designed to assist small and medium-sized enterprises (SMEs), particularly those in the electro- deposition paint (EDP) sector. The purpose of this system is to simplify the stock management process and minimize human error in raw material recording. The primary focus of this project is the implementation of a Demand Forecasting feature designed to predict future raw material needs based on historical data from the past three months. The Demand Forecasting feature we will develop will use linear regression methods and monthly data processing to estimate the amount of material needed for that month, thereby preventing stock shortages or surpluses. Additionally, this web application includes a notification feature aimed at alerting users if any raw material falls below the minimum threshold. In addition to the main features mentioned above, this project includes features specific to each role. The author has developed features for the management and driver roles, The management role can access reports of all activities, such as item usage, delivery progress, and other operational data. Meanwhile, the driver role is responsible for updating the delivery note status for example, changing it from 'Outgoing' to 'Delivered'.. The web application's UI will be developed using Bootstrap v5.3, jQuery, DataTables, and SweetAlert, with MySQL used for data storage. All development for this project was carried out using Visual Studio Code (VSCode), with a local server using XAMPP. Testing results show that the prediction feature operates as intended, with a low error rate and reliable results in assisting users with the raw material procurement planning process. This system effectively assists in the planning and availability of raw materials for these SMEs. 2025-01-01T00:00:00Z http://repository.president.ac.id/xmlui/handle/123456789/13311 SMART MANUFACTURING EXECUTION SYSTEM: REVOLUTIONIZING INDUSTRY WITH IOT AND AI (MODULE: IOT AND TELEGRAM ALERT) Nainggolan, Kevin Lavpienji Industrial environments need reliable and fast data monitoring to maintain operational efficiency and safety. This project takes on the task of constructing a secure Industrial Internet of Things (IIoT) pipeline to create an end-to-end data collecting and visualization system. My work is centered on the system's basic integration and data processing elements. The approach entailed establishing Node-RED to communicate directly with a Programmable Logic Controller (PLC) and collect vital process data. This data was then published to a MQTT broker, and the communication channel was encrypted with TLS certificates created by Step-CA to assure data integrity and secrecy. As a result, I created a service to listen to MQTT topics, parse the incoming data, and store it in a TimescaleDB database for efficient time-series analysis. For data presentation, I built a Flask-based web application with two primary functions: it embeds a Grafana dashboard that queries TimescaleDB for historical data visualization, and it also uses a WebSocket connection to display live data directly from the MQTT stream, providing an immediate, real-time view of the industrial process. The final product is a safe and fully functional monitoring tool that later can send the user both live and historical data. This shows that industrial automation hardware can be successfully combined with modern open-source IoT technology. 2025-01-01T00:00:00Z http://repository.president.ac.id/xmlui/handle/123456789/13310 IMPLEMENTATION OF FACE RECOGNITION TECHNOLOGY IN AN ATTENDANCE SYSTEM BASED ON RASPBERRY PI AND WEB INTERFACE Ndraha, Charles Paskah Ifohaga This study discusses the design and implementation of an automatic attendance system using facial recognition technology running on the Raspberry Pi 4B platform. Manual attendance methods are often considered inefficient, prone to manipulation, and time-consuming, especially in academic and office environments. By leveraging biometric technology and low-cost edge computing devices, this study aims to provide a practical, real-time, and contactless solution for recording attendance. The system uses a USB camera connected to the Raspberry Pi to capture facial images. Face detection is performed using OpenCV with the Haar Cascade classifier, while face recognition is carried out using the Local Binary Pattern Histogram (LBPH) algorithm due to its simplicity and efficiency on devices with limited computational power. Once the face is recognized, the system automatically records the name and attendance time in a structured .csv file. To enhance usability and accessibility, the system is also equipped with a lightweight Flask-based web interface that allows administrators to view, manage, and download attendance data via a browser. The entire process—from face detection, recognition, recording, to web display— is run locally on the Raspberry Pi without requiring a cloud connection, demonstrating that artificial intelligence technology can be directly implemented on edge devices. Testing was conducted under various lighting conditions and facial angles to evaluate the system's accuracy and speed. Results show that the system can provide reliable recognition rates and efficient processing times for small-scale applications such as classrooms or offices. This research contributes to the development of intelligent systems in education and organizational settings by providing an affordable, scalable, and user-friendly attendance solution. Further development could include integration with a centralized database, algorithm improvements using deep learning, and multi- device synchronization support. Overall, the implementation of a face recognition-based attendance system on Raspberry Pi, equipped with a web interface and structured data storage, has proven to be a practical and efficient tool for attendance automation with minimal human intervention and high accessibility. 2025-01-01T00:00:00Z