This project introduces an AI-powered teacher-student ecosystem that revolutionizes assessment creation, personalization, and evaluation in education. The core innovation lies in leveraging a Hybrid AI approach that integrates Large Language Models (LLMs) for question generation and Smaller Language Models (SLMs) like BERT for intelligent question classification. The system aligns all assessments with Bloom’s Taxonomy, ensuring cognitive diversity across six levels: Remember, Understand, Apply, Analyze, Evaluate, and Create.

Traditional assessment methods are manual, inconsistent, and lack adaptability. Our solution addresses these limitations by automating the end-to-end process—from generating high-quality, taxonomy-based questions to grading and performance analytics. A rule-based system manages test scheduling, scoring logic, and personalized reporting.

Students benefit from adaptive learning paths based on their uploaded study materials, enabling continuous self-evaluation and targeted improvement. The platform dynamically adjusts question difficulty and focus areas to match individual performance and learning gaps.

Technically, the system combines natural language understanding, text generation, classification algorithms, and rule-based logic into a unified, scalable architecture. This hybrid design ensures both flexibility and reliability, optimizing for both accuracy and performance.

The potential impact is transformative:

Educators save significant time and ensure cognitive balance in assessments.
Institutions benefit from standardized, scalable evaluations.
Students receive personalized feedback and critical thinking development.

By aligning AI capabilities with pedagogical best practices, this project paves the way for data-driven, equitable, and future-ready education.

The idea for this project emerged from firsthand observation of the challenges educators and students face in modern classrooms. Teachers invest significant time manually creating and grading assessments, often struggling to ensure that questions align with cognitive learning objectives such as those defined in Bloom’s Taxonomy. Meanwhile, students frequently receive generic, non-personalized assessments that don’t address their unique learning needs or gaps in understanding.
We recognized a gap: while educational content has advanced, the assessment process remains largely static and inefficient. Inspired by the recent advancements in AI, particularly large and small language models, we saw an opportunity to build a smart, automated ecosystem that bridges this divide. Our goal was to empower educators with tools that reduce workload and enhance assessment quality, while also giving students meaningful, adaptive learning experiences.
The core hypothesis behind our project is: combining generative AI (LLMs) with classification AI (SLMs), anchored in Bloom’s Taxonomy and guided by rule-based automation, can create a scalable, effective, and personalized assessment system.

We believe this approach will succeed because:

AI has matured to the point where it can reliably generate and categorize questions across learning levels.
Automation can handle repetitive tasks like grading and scheduling, freeing educators to focus on instruction.
Personalization is increasingly expected in digital education; adaptive assessments directly meet this need.

This project aligns with the evolving landscape of education, where scalable, data-driven tools are essential for delivering quality, inclusive, and future-ready learning.

Our solution leverages a Hybrid AI architecture combining Large Language Models (LLMs) like GPT for automated question generation and Small Language Models (SLMs) such as BERT for categorizing questions according to Bloom’s Taxonomy. These technologies will be integrated into a web-based platform with a user-friendly dashboard for educators and students. Rule-based systems will manage test scheduling, grading logic, and report generation, ensuring seamless automation.

To build this platform, we require:

Access to LLM and SLM APIs (e.g., OpenAI, gemini)
Cloud infrastructure (e.g., GCP) for model deployment and data storage
Frontend and backend development resources (React,Python)
Education domain experts for content validation and feedback

We plan to validate market demand through:

Pilot programs with schools and coaching centers
Surveys and interviews with educators and students
Early beta testing and feedback loops

Core functionalities of the platform include:

AI-powered question generation aligned with Bloom’s levels
Automated classification of cognitive complexity
Personalized assessments based on uploaded learning material
Auto-grading and performance analytics
Adaptive practice modules for students

To ensure a positive user experience, we have implemented:

Intuitive UI/UX design with minimal learning curve
Customization options for educators and learners
Continuous user feedback integration
Data privacy and accessibility compliance

Performance metrics to evaluate effectiveness:

Accuracy of AI-generated question alignment with Bloom’s levels
Reduction in time spent on assessment creation
Improvement in student performance over time
User satisfaction and platform engagement rates

Our approach balances technical innovation with educational relevance, ensuring both feasibility and high-impact functionality.

Our system architecture is designed around a modular, scalable AI-driven ecosystem that integrates state-of-the-art technologies for seamless automation and performance. The functional architecture consists of four core components: frontend interface, backend processing, AI orchestration, and data management.

Frontend:
Developed using React, the UI provides intuitive dashboards for both educators and students, enabling question generation, test creation, personalized practice, and performance review.

Backend & AI Orchestration:
At the core, we use Gemini as the LLM for natural language generation to create diverse, contextually relevant, and Bloom’s Taxonomy-aligned questions. LangChain and LangGraph serve as the orchestration layer to manage conversational context, prompt routing, and agent logic for adaptive assessments and iterative feedback. Rule-based engines handle test scheduling, grading logic, and automated feedback generation.

Database:
We use MongoDB as the primary data store for user profiles, assessment metadata, student responses, and performance analytics. The schema is optimized for fast querying and scalability.

Performance Metrics:

Accuracy of Bloom’s-level classification
Average question generation and grading latency
Student performance improvement over iterations
Educator time saved (vs manual workflows)
Platform engagement and satisfaction rates

This implementation plan ensures a robust, flexible, and data-driven solution that tightly integrates advanced AI technologies with real-world educational needs.

Our project introduces a novel AI-powered ecosystem that transforms how assessments are created, delivered, and evaluated in education. While traditional platforms focus on digitizing static tests, our solution takes a fundamentally innovative approach by combining generative AI (Gemini) with orchestrated reasoning (LangChain + LangGraph) and a taxonomy-aligned classification engine, all integrated into a seamless user experience.

The creative use of Bloom’s Taxonomy as a backbone for question generation and categorization sets our platform apart. Unlike conventional systems that generate generic questions, our solution ensures that each item targets a specific cognitive level—fostering deeper learning and critical thinking. This structured, yet dynamic, framework allows educators to create balanced and meaningful assessments effortlessly.

Our platform's ability to personalize learning experiences using student-uploaded content and adaptive feedback is another unique innovation. By analyzing individual learning gaps and automatically generating relevant practice questions, the system not only saves educators time but also empowers students to take ownership of their progress.

The use of LangGraph for agentic workflows adds another layer of creativity. It enables multi-step reasoning and adaptive dialogue with students, allowing for deeper interaction and real-time scaffolding, far beyond static quizzes.

From a user experience perspective, we’ve reimagined assessment as a collaborative and intelligent process rather than a one-time test. Features like automated grading, real-time analytics, and performance-based content adaptation create a feedback-rich loop that continuously supports learner growth.

By tightly integrating advanced AI with proven educational frameworks, this project exemplifies innovation in both technology and pedagogy—solving real-world classroom problems with scalable, intelligent, and personalized solutions.

Our platform is architected for scalability from the ground up, leveraging cloud-native technologies and modular AI orchestration to handle growing user demand without performance degradation. We use serverless and containerized deployment strategies (via AWS or GCP) to dynamically scale compute resources based on load. Gemini’s API-based architecture and LangChain/LangGraph orchestration allow for parallel processing of assessments and real-time personalization at scale. MongoDB’s flexible schema design ensures rapid, scalable data storage and retrieval across large datasets.

To address potential bottlenecks, we will:

Implement intelligent request throttling and load balancing for AI inference calls.
Use caching layers for frequently accessed data (e.g., previously generated questions or student history).
Continuously monitor system performance using observability tools like Prometheus and Grafana.

For sustainability, we focus on:

Cloud efficiency: Using resource-optimized LLMs and inference endpoints to reduce energy consumption.
Model tuning: Incorporating smaller, fine-tuned SLMs for classification tasks to reduce compute load.
Modular updates: Ensuring that individual components (e.g., feedback logic or UI) can be updated without overhauling the entire system—prolonging the platform’s lifecycle.

To maintain long-term user engagement, we provide:

Gamified learning experiences and badges based on progress.
Real-time feedback and personalized recommendations that evolve with user performance.
Tools for educators to customize and share assessments across institutions.

Our solution directly addresses key social issues in education—inequity, lack of personalization, and limited access to quality assessment tools—by making intelligent learning support universally available. Many students, particularly in under-resourced communities, lack personalized academic support. Similarly, educators are burdened by manual tasks that hinder their ability to focus on meaningful instruction.

By automating assessment creation and aligning it with Bloom’s Taxonomy, we ensure that students at all cognitive levels receive balanced, skill-targeted learning opportunities. Our AI-powered system supports self-paced learning, allowing students to progress based on their individual needs and strengths. This is especially beneficial for students with learning differences or those from diverse language and socio-economic backgrounds.

Our platform also empowers teachers in underserved schools by removing the technical and time barriers involved in designing high-quality assessments. With accessible tools and analytics, educators can provide timely interventions and track growth—enhancing educational outcomes across demographics.

We align with broader goals of equity, inclusion, and lifelong learning by:

Supporting multilingual capabilities for diverse student populations
Offering a freemium model to ensure access in low-income regions
Designing for accessibility (WCAG-compliant interface and low-bandwidth optimization)

Social impact metrics include:

Increase in student performance and confidence (via pre-/post-assessment data)
Reduction in teacher workload (measured by time saved)
Engagement rates among students from marginalized communities
User feedback and qualitative insights from pilot implementations

To remain responsive to evolving community needs, we will:

Partner with educators, NGOs, and policymakers for ongoing feedback
Maintain a live feedback loop and feature suggestion board
Continually adapt based on emerging educational standards and inclusion goals

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.