A Review of AI-Based Job Portal Systems for Intelligent Job Matching and Recruitment

The growth of digital recruitment has created large volumes of job postings, candidate profiles, resumes, applications, and interaction data. Conventional job portals mainly support job publication, keyword search, filtering, application submission, and employer-side candidate management. Although these functions improve access to employment opportunities, they do not always solve the core matching problem: identifying the most relevant job for a candidate and the most suitable candidate for an employer. Artificial Intelligence (AI), Machine Learning (ML), Natural Language Processing (NLP), and recommender-system techniques provide opportunities to improve this process by understanding skills, experience, qualifications, job descriptions, preferences, and historical interactions. This review paper examines research on job recommender systems, e-recruitment recommendation, semantic job matching, resume screening, deep-learning-based talent recommendation, and algorithmic fairness. The literature indicates a progression from content-based and collaborative filtering toward hybrid, semantic, and deep-learning approaches. Recent studies further emphasize explainability, fairness, semantic embeddings, and large-language-model-assisted recruitment. Based on the reviewed literature, this paper proposes an AI-enabled job portal architecture combining a React Native mobile client, Spring Boot backend, MySQL database, and Python-based AI services. The proposed architecture supports role-based access for job seekers, employers, and administrators; resume processing; skill extraction; semantic candidate-job matching; personalized job recommendations; application tracking; employer notifications; and chatbot assistance. The paper identifies major research gaps involving cold-start users, limited public recruitment datasets, explainability, fairness, dynamic job preferences, and integration of multiple recruitment functions into a practical end-to-end platform. Finally, evaluation metrics and an experimental methodology are proposed for assessing recommendation quality, classification performance, response time, usability, and fairness.