System Architecture
The system architecture showcases the complete workflow from data collection through Kaggle to final job recommendations.
AI-powered career matching platform using BERT and contextual alignment to match resumes with job postings, providing skill-gap analysis and match percentages for the Malaysian job market.
This project implements a hybrid algorithm combining Content-Based Filtering (CBF) and Natural Language Processing (NLP) techniques, enhanced with Bidirectional Encoder Representations from Transformers (BERT) for contextual understanding. The system uses Term Frequency-Inverse Document Frequency (TF-IDF) for feature extraction and Cosine Similarity for matching user profiles with job descriptions.
BERT fine-tuning adapts to the Malaysian job market's linguistic nuances for accurate job matching.
Cosine Similarity ranks jobs based on relevance to user profiles with 88% accuracy.
Efficient processing of large datasets with 59,306 job postings analyzed.
Identifies missing skills and provides recommendations for career advancement.
BERT embeddings outperform traditional keyword-based systems.
Aligns recommendations with career goals and existing skill sets.
The system architecture showcases the complete workflow from data collection through Kaggle to final job recommendations.
Quantifies the importance of words in resumes and job descriptions for feature extraction.
Captures semantic relationships in text for deeper contextual alignment and understanding.
Measures similarity between user profiles and job postings for accurate matching.
Identifies key entities like skills, job titles, and locations from text data.
Resumes and job descriptions are parsed using spaCy and PDFMiner for accurate data extraction.
Tokenization, lemmatization, and stopword removal standardize text for analysis.
TF-IDF and BERT convert text into numerical vectors for machine learning.
Fine-tuned BERT model categorizes jobs into Entry, Mid, or Senior levels.
Cosine Similarity generates a ranked list of matched jobs with percentages.
User receives personalized job recommendations with skill gap analysis.
The web-based system provides an intuitive interface for resume upload, profile management, and job recommendations.
Achieved 88% accuracy in job level classification, outperforming traditional systems.
Reduces job search time by 70%, helping users find relevant positions faster.
Recognized at WINSTEM 2025, Borneo Innovation Festival, and COMMAX 2025.
Helps employers find qualified candidates efficiently through semantic matching.
Guides students in career planning aligned with their academic profiles.
Analyzes extensive dataset from JobStreet Malaysia for comprehensive coverage.
This AI-powered Career Recommendation System contributes to socio-economic growth by enhancing career accessibility and workforce efficiency, aligning with United Nations Sustainable Development Goals.
Enhances career readiness by providing personalized job recommendations that align students' skills with market demands, bridging the gap between education and employment. Identifies skill gaps and suggests relevant upskilling opportunities, supporting lifelong learning.
Reduces youth unemployment by guiding job seekers toward suitable roles, minimizing mismatches and underemployment. Boosts workforce productivity by helping employers find qualified candidates efficiently, strengthening Malaysia's labor market.