Edunavigator An AI Powered Career Counseling & College Recommendation Platform Integrating on Premise Large Language Models Adaptive Assessment & Real-World Cutoff Data

India's competitive entrance examination ecosystem creates a persistent advisory bottleneck: millions of students annually make high-stakes academic decisions with minimal personalized guidance. This paper presents EduNavigator, a full- stack, AI-driven web platform built with Node.js, Express.js, MongoDB, and EJS that integrates three AI providers — a locally deployed Llama 3 model via Ollama, the Groq cloud inference API, and the Google Gemini API — to deliver intelligent career counseling, visual roadmap generation, adaptive skill assessment, and score-based college recommendations. Unlike cloud-only solutions, EduNavigator processes latency-sensitive inference on- premise, guaranteeing student data privacy and eliminating recurring AI expenditure. The platform unifies eight feature modules: an AI chatbot with PDF resume analysis and Mer- maid.js visual career roadmaps; an AI career trend predictor; a 60-mark adaptive quiz engine with multi-dimensional radar analytics; a CET/JEE college recommender backed by a verified seven-year cutoff dataset (2018–2025); a real-time WebSocket leaderboard; a scholarship finder; a project-based learning tracker; and a Gemini-powered study-abroad advisor. Security is enforced through Passport.js local and Google OAuth 2.0 authentication, Helmet.js HTTP hardening, Express rate limiting, and MongoDB sanitization. A user acceptance study with twenty student participants yielded an 88% AI counseling satisfaction rate (average Likert score 4.1/5), 85% college recommendation relevance, and a 90% platform recommendation rate. This work contributes a reproducible, open-source reference architecture for scalable, privacy-preserving AI-driven academic advisory in resource- constrained educational institutions.