The AI-Powered Tutoring System is developed to address the persistent challenges of limited personalized

learning support and cognitive overload that students experience when managing and revising extensive

architecture comprising React.js for the frontend, Node.js and Express for the backend, and PostgreSQL as the

database. Guided by the Agile methodology, the development process was structured into iterative sprints

encompassing key phases such as planning, design, development, and testing. The system integrates the

DeepSeek API for natural language processing, enabling the platform to provide summarized lecture notes and

structured quizzes tailored to each uploaded document. It also includes user authentication, file handling,

progress tracking, and a personalized library, allowing users to manage their learning resources effectively. A

user acceptance test (UAT) with 31 undergraduate students was conducted using a five-point Likert scale

questionnaire. The UAT results show that perceived usefulness, perceived ease of use, capability,

trustworthiness, attitude toward the system, and behavioral intention to use had a high acceptance rate.

Theoretically, this work contributes to the advancement of AI-driven educational technology by integrating

Many students today face strong pressure to succeed in exams, especially when they need to study a large

amount of theory from different subjects. Traditional study methods often demand significant time and effort,

especially when students struggle to understand complex topics or memorize essential information. Common

methods like memorization and relying on disorganized digital files often lead to low retention and poor

understanding. Students frequently struggle to organize their notes effectively, identify key points within

lengthy documents, and assess their knowledge in a clear and structured way. One major problem is the limited

time available for revision, especially during final exam periods. As a result, students frequently resort to less

effective methods such as rote memorization or relying on online materials that are not aligned with their

specific course content. This can hinder deep learning and long-term retention.

A recent study highlighted various factors influencing students’ learning performance and emotional well-

both opportunities and challenges in adopting AI in education, calling for research to investigate how learning

objects are to be used in personalized and adaptive learning.

In response to these challenges, this study aims to design, develop, and evaluate an AI-Powered Tutoring

System that automatically summarizes academic materials and generates quizzes to support students’ self-

directed learning. This web-based application is designed to assist students in their independent learning by

providing smart tools for organizing, analyzing, and processing study materials. Users can upload academic

resources such as PDFs or text files, which are processed by the system’s artificial intelligence engine to

generate concise summaries and create personalized quiz questions tailored to the content. The system is built

using modern web technologies, including React for the user interface, Node.js for backend functionality, and

PostgreSQL for data management. The platform includes performance tracking features that allow students to

Recent studies highlight the growing role of artificial intelligence in education, particularly in intelligent

tutoring systems, automated notes summarization, and quiz generation. Studies from [3], [4] report that AI-

driven platforms can improve learning outcomes by enhancing engagement, providing adaptive feedback, and

simplifying access to key concepts. These findings reflect the growing demand for educational technologies

that accommodate diverse learning styles and help students manage large volumes of academic content

efficiently. Moreover, [5] explores AI-based approaches for students to learn Python through quizzes and

discussions. The findings indicate that generative AI serves as an effective resource for educational

advancement, as it can lower learning anxiety and increase interest in learning.

Table I shows the comparison between existing systems and the proposed AI-Powered Tutoring. System.

but it is primarily designed for course creators and lacks automated summarization or learner-centered

adaptation. Overall, most existing systems do not holistically support constructivism and self-regulated

learning principles. Constructivist theory emphasizes the active role of learners in constructing meaning [10],

while self-regulated learning highlights the importance of autonomy, goal-setting, and reflection [11].

Compared to these platforms, the proposed AI-Powered Tutoring System uniquely combines automated

summarization and quiz generation with user-uploaded academic materials (PDF, TXT, PPT). By integrating

the DeepSeek API for natural language processing and a full-stack web architecture (React, Node.js, Express,

PostgreSQL), the system allows students to generate structured summaries and quizzes from their own notes

while tracking performance over time. This high level of personalization and content control addresses gaps in

current solutions and supports more effective, self-directed learning.

The project adopts an Agile methodology with the Scrum framework. As highlighted by [12]Agile

methodology supports iterative development, continuous feedback, and adaptability, which are the key features

that are essential for the successful development of complex systems such as AI-powered tutoring platforms.

manageable tasks that could be implemented over several Agile sprints.

Once planning was complete, the system architecture and user interface were designed. Fig. 1 illustrates the

system architecture, highlighting the interaction between the frontend, backend, database, and external AI

processing layer. The system is designed based on client-server architecture, which delineates responsibilities

between two main components: the client and the server. This architectural model ensures modularity,

simplifies maintenance, and supports future scalability. It also enables secure and efficient interaction between

system components during data processing and retrieval.

Fig. 1 System Architecture

UI wireframes were created to visualize the layout of major components, including the dashboard, upload

Fig. 3 (a) Generate Quiz, (b) Select Quiz Difficulty, (c) Quiz question

23 items, as presented in Table II. A 5-point Likert scale questionnaire items were adapted from [13], [14]. The

questionnaire items were divided into six constructs named perceived ease of use, usefulness, capability,

Table2 Uat Questionnaire Items

The findings demonstrate that the proposed AI-powered system can enhance student engagement, autonomy,

and self-directed learning by providing personalized summaries and quizzes. For practical application,

educators and instructional designers can integrate such systems into blended and online learning

the mobile responsiveness is functional but not fully optimized for smaller screens, which could affect

usability on smartphones. A dedicated mobile layout or even a Progressive Web App (PWA) version could

allow students to use the system on smartphones more comfortably, especially during quick revision sessions.

Lastly, the system relies on internet connectivity and the availability of the DeepSeek API, meaning it cannot

operate offline. Implementing offline functionality using service workers and local storage could allow users to