Bilimge: Design and Implementation of an AI-Assisted Platform for University Admissions in Kazakhstan

Bilimge is a web application integrating artificial intelligence methods, designed to provide exam preparation and admission support for university applicants in Kazakhstan. Through features such as personalized learning path planning, intelligent recommendations, and automated guidance, it addresses challenges faced by students in the University Entrance Exam (UNT), including information asymmetry, scattered resources, and lack of personalized guidance, helping students enter their ideal universities while promoting educational equity.

In recent years, Kazakhstan has vigorously developed higher education and is home to well-known institutions such as Al-Farabi Kazakh National University. Each year, a large number of high school graduates face the UNT exam, but exam preparation has four major challenges:

1. Information asymmetry: Difficulty in obtaining university major information, admission criteria, and employment prospects
2. Scattered preparation resources: Lack of systematic integration of learning materials
3. Lack of personalized guidance: Traditional models cannot provide customized tutoring for weaknesses
4. Psychological pressure: College entrance exam anxiety affects efficiency The Bilimge project is a comprehensive AI-assisted platform developed to address these issues.

User Roles

The system serves three types of users:

• Applicants: Learning profile management, intelligent path planning, mock exams, university recommendations, progress tracking
• Content administrators: Resource upload management, question bank maintenance, system configuration
• Administrators: User management, data analysis, system settings

AI Core Functions

1. Personalized learning path: Generate customized plans based on knowledge graphs, ability assessments, and reinforcement learning
2. Intelligent question bank and adaptive testing: IRT difficulty modeling, dynamic question adjustment, wrong answer analysis, and similar question generation
3. University and major recommendations: Multi-dimensional matching, admission probability prediction, employment prospect analysis
4. NLP Q&A assistant: Admission policy consultation, subject Q&A, psychological counseling, multi-language support (Kazakh/Russian/English)

Frontend

React.js + Redux + Material-UI + Chart.js

Backend

Node.js/Express + PostgreSQL + Redis + MongoDB

AI/ML Technology

• Recommendation system: Collaborative filtering, matrix factorization, neural collaborative filtering
• NLP: Multilingual BERT, text classification, RAG architecture
• Knowledge tracking: BKT, DKT, Transformer-based
• Adaptive learning: Multi-armed bandit, contextual bandit, reinforcement learning

Deployment and Operation

Docker containerization + AWS cloud services + CI/CD pipeline + Prometheus/Grafana monitoring

Pilot Results

• Improved learning efficiency: Average study time reduced by 20%, scores increased by 15%
• Clearer goals: 85% of students have a clearer understanding of university choices
• Reduced pressure: 70% of students believe exam anxiety is alleviated
• Increased resource utilization

User Feedback

Students:

• Intelligent learning plans make daily learning goals clear
• Wrong answer analysis helps identify weak areas
• University recommendations help discover unknown good majors

Teachers:

• Learning situation reports help understand students
• Automated test scoring saves time

Multilingual Support

Challenge: Limited Kazakh NLP resources, maintaining context during language switching Solution: mBERT/XLM-R base models + Kazakh corpus pre-training + language-agnostic features

Data Sparsity (Cold Start)

Solution: Content-based recommendation + interest and ability assessment questionnaire + similar user group data

Model Interpretability

Solution: Recommendation explanation notes + knowledge graph visualization + SHAP/LIME analysis

Future Plans

• Short-term: Expand question bank, optimize mobile version, add interactive elements
• Mid-term: Expand to Central Asian countries, VR immersive learning, parent-side application
• Long-term: Become a leading AI education platform in Central Asia, establish a big data center, promote policy standards

Social Value

• Narrow urban-rural gap: Students in remote areas gain equal access to resources
• Reduce costs: Replace expensive cram schools
• Empower autonomous learning: Cultivate lifelong skills such as goal setting and time management

Bilimge demonstrates the great potential of AI in the education field. It provides support for Kazakh applicants through features like personalized learning and promotes educational equity. In the future, similar systems are expected to be implemented in more countries, helping global learners prepare for the future.