AI-based Job Optimizer: An Intelligent Recruitment Platform using NLP, LLMs, and Hybrid Recommender Systems

This paper describes an AI-powered, dual-dashboard recruitment platform that integrates several LLMs (ChatGPT, Gemini, and Ollama) with semantic vector-based candidate-job matching. This system uses BERT-style embeddings and Supabase's PostgreSQL backend with a pgvector extension to perform real-time and scalable candidate-job matching. It provides user-selectable AI models for flexible semantic analyses that range from contextual comprehension through ChatGPT to multimodal processing with Gemini and even privacyfocused local deployment with Ollama. These are combined with multi-factor scoring that integrates skill overlap, experience alignment, location compatibility, and salary expectations. Experimental evaluation on a dataset of 500 candidates and 500 job applications with relevant job postings and candidate profiles yields superior matching accuracy compared with keywordbased methods, with Krippendorff's alpha = 0.6287 vs. 0.2262. The ACID-compliant relational database of Supabase, integrated with auto-generated REST APIs, RLS policies, and real-time synchronization over WebSockets, ensures the secure, highly scalable operation of this application. Matching queries are returned within sub-200ms responses in the system for responsive user experiences during peak operations. Unlike traditional keyword-based systems or single-model AI platforms, this work demonstrates that multi-model flexibility with semantic embeddings and real-time infrastructure brings significant value in improving hiring efficiency and user experience without compromising data privacy and compliance with security regulations.