Higher Education Category Submission

517 – Personalized Learning Assistant: AI-Powered MCQ-Based Adaptive Assessment

The Personalized Learning Assistant (PLA) is an AI-driven educational platform designed to assess and enhance students' understanding through adaptive MCQ-based tests. The system identifies students' weak sub-topics based on their responses and dynamically generates personalized quizzes to improve their knowledge.
Key Innovations:
AI-Powered Question Generation: Uses fine-tuned LLMs to generate MCQs from text passages.
Adaptive Learning: Adjusts difficulty levels dynamically based on student performance.
Real-Time Feedback & Analytics: Tracks progress and provides insights into weak areas.
Gamification Elements: An engaging React-based UI to enhance motivation and retention.

Technical Principles:
Backend: Hosted on Kaggle using FastAPI, handling question generation, evaluation, and analytics.
Frontend: Built with React (Bootstrap), calling APIs in real-time for MCQ generation and performance evaluation.
LLMs:
Custom fine-tuned models trained on our own NCERT-based dataset to generate curriculum-aligned MCQs.
AI classifies MCQs into granular sub-topics for a targeted learning experience.

Deployment & Accessibility:
The AI model runs on Hugging Face Spaces, integrated with a Gradio interface for interactive testing.
Designed to be lightweight and scalable, supporting multiple subjects.

Potential Impact:
Teacher-Assisted Learning: PLA is not a replacement for teachers but an AI-powered assistant to help them identify students' strengths and weaknesses efficiently.
Personalized Education: Adapts to each student’s learning pace, ensuring a tailored experience.
Scalability Across Subjects & Grades: The system can be expanded to multiple subjects beyond the initial NCERT dataset.
Bridging Learning Gaps: Provides targeted intervention, helping students focus on specific weak areas and reinforcing conceptual understanding.

The motivation for this project first came when Harvard University announced the introduction of an AI teacher for their CS50 course. This led us to ask:
"How can AI transform education?"

As we explored this idea, we came across research from National Taichung University on the Taiwan Adaptive Learning Platform (TALP). This study demonstrated how AI-driven platforms could tailor education to individual students' needs, reinforcing our belief that AI could revolutionize learning. Inspired by this, we decided to develop our own Personalized Learning Assistant (PLA).
PLA is not designed to replace teachers but to assist them, providing a better and more immersive learning experience. The key challenge in education today is personalization—a single teacher managing a classroom of 30-40 students cannot easily track each student's strengths and weaknesses. With PLA, teachers can identify sub-topics where students struggle and intervene effectively.
Our hypothesis is that AI-driven personalized assessments can enhance learning outcomes by identifying weak areas and adapting the content accordingly. PLA aims to:
Predict the sub-topics in which a student is weak.
Generate adaptive MCQs to help students strengthen those areas.
Provide teachers with data-driven insights to support targeted interventions.
By leveraging AI and real-time analytics, PLA enables individualized learning paths that would otherwise be difficult to implement in large classrooms. This approach fosters engagement, retention, and improvement in student performance, making education truly personalized and effective.

-

The development of the Personalized Learning Assistant has been progressing steadily, with several key technical components successfully implemented.
The foundation of the project lies in the custom dataset creation, which has been meticulously curated using NCERT-based content. This dataset plays a crucial role in both MCQ generation and sub-topic classification, ensuring that the generated questions align with educational standards. A fine-tuned Large Language Model (LLM), specifically trained on this NCERT dataset, has been successfully developed for generating high-quality MCQs. Alongside this, an AI-driven classification model has been built to accurately categorize sub-topics within educational content, aiding in personalized learning experiences.
On the backend, the system is powered by FastAPI, a robust and efficient framework, which has been successfully deployed on Kaggle to handle requests and ensure seamless interaction between different modules. For the front end, a dynamic and user-friendly interface is being developed using React and Bootstrap. While it is currently in progress, once completed, it will offer an engaging and intuitive experience for users.
To enhance user engagement and provide meaningful insights, real-time evaluation is being integrated, allowing learners to receive instant feedback on their responses. This feature, which involves API integration for real-time analysis, is actively under development. Additionally, to make learning more interactive and motivating, gamification features such as progress tracking, interactive UI elements, and achievement-based incentives are planned for future implementation.
With most of the core functionalities successfully completed and others in the pipeline, the project is steadily moving towards its goal of providing an intelligent, adaptive, and engaging learning experience.

The Personalized Learning Assistant (PLA) introduces an innovative approach to AI-driven adaptive learning, specifically designed to assist educators rather than replace them. Unlike traditional learning platforms that provide static quizzes, PLA dynamically generates MCQs in real time, ensuring a truly personalized learning experience tailored to each student’s needs.
Key Innovations:
Custom LLM Fine-Tuned on NCERT Data – Instead of relying on generic pre-trained models, PLA uses a fine-tuned language model trained on structured educational content, ensuring accurate and curriculum-aligned MCQ generation.

Granular Sub-Topic Classification – Most educational tools assess broad topics, but PLA goes deeper by classifying student performance at the sub-topic level, helping teachers pinpoint specific weaknesses.

Adaptive Difficulty & Smart Remediation – PLA adjusts the difficulty of questions in real-time based on performance and provides remedial questions for incorrect answers, ensuring a structured learning progression.

Gamified Learning Experience – Inspired by interactive learning apps, PLA integrates an engaging UI and real-time feedback, making the learning experience immersive for students.

Creativity in Addressing Challenges
Bridging the Personalization Gap in Large Classrooms: Teachers often struggle to provide individual attention in classrooms with 30-40 students. PLA acts as a teaching assistant, highlighting learning gaps and allowing educators to provide targeted support.

Scalability & Versatility: PLA is designed to scale across multiple subjects and educational levels, making it adaptable for different curricula.

Real-Time Performance Insights: Unlike conventional assessment platforms, PLA gives instant analytics on a student’s strengths and weaknesses, empowering both students and teachers to take corrective actions immediately.

The Personalized Learning Assistant (PLA) is designed to handle increasing user demand efficiently through the following strategies:
Cloud-Based Infrastructure – The backend, built using FastAPI and hosted on Kaggle, can be migrated to cloud services like AWS or Google Cloud for higher scalability. This ensures seamless performance even with a growing number of users.

Optimized API Calls – The system efficiently caches frequently generated MCQs and minimizes redundant API requests to reduce server load and improve response times.

Containerization & Load Balancing – By using Docker and Kubernetes, PLA can scale automatically based on demand, ensuring smooth operation during peak usage.

Edge AI for Faster Processing – Future iterations will leverage on-device inference for MCQ generation to reduce dependency on cloud processing and enhance accessibility in low-bandwidth areas.

Sustainability Strategies
Efficient Resource Utilization: PLA is optimized to use lightweight LLM models where possible, reducing energy consumption compared to larger AI models.

Educational Equity: The platform is designed to work efficiently on low-end devices and in regions with limited internet access, promoting inclusive education.

Long-Term Engagement:
Gamification (e.g., streaks, achievements, and an interactive mascot) keeps students motivated.
AI-driven personalization ensures that students always receive content suited to their needs, keeping them engaged.
Adaptability: PLA will continuously learn from user interactions, refining its AI model to align with evolving educational trends and curriculum updates.

The Personalized Learning Assistant (PLA) is designed to bridge educational gaps, making learning more accessible and effective for students of all backgrounds. It supports equity and inclusion in the following ways:
Personalized Support for Every Student – Traditional classrooms often struggle with one-size-fits-all learning. PLA ensures every student gets tailored support, helping those who might otherwise be left behind.

Inclusive Education – The tool works across different subjects, education levels, and languages, ensuring students from diverse backgrounds can benefit. Future iterations will support regional languages to reach underserved communities.
Supporting Teachers, Not Replacing Them – PLA assists teachers by providing real-time insights into student performance, helping them focus on students who need the most attention.

Alignment with Broader Social Goals
Democratizing Education: PLA ensures quality learning is accessible, reducing disparities between students in urban and rural areas.
Enhancing Digital Literacy: By integrating AI-powered learning tools into the classroom, PLA fosters early adoption of technology, preparing students for future careers.

Measuring Social Impact
PLA’s impact will be assessed using:
Engagement Metrics – Number of active users, time spent learning, and improvement trends.
Learning Outcomes – Improvement in weak sub-topics, as measured through pre-and post-test scores.
Teacher Adoption Rate – How many educators use PLA to enhance their teaching methods.
User Feedback & Adaptation – Regular feedback loops with students and teachers will help refine PLA’s effectiveness.

Personal Information Collection Statement (PICS):
1. The personal data collected in this form will be used for activity-organizing, record keeping and reporting only. The collected personal data will be purged within 6 years after the event.
2. Please note that it is obligatory to provide the personal data required.
3. Your personal data collected will be kept by the LTTC and will not be transferred to outside parties.
4. You have the right to request access to and correction of information held by us about you. If you wish to access or correct your personal data, please contact our staff at lttc@eduhk.hk.
5. The University’s Privacy Policy Statement can be access at https://www.eduhk.hk/en/privacy-policy.

• I have read and agree to the competition rules and privacy policy.