Every line of this project began with a single observation: the students who most need expert admissions guidance are exactly the ones who can least afford it.
I am an international applicant. Over two years I watched peers navigate the US admissions process. Students with access to private counselors — paying $3,000 to $50,000 per application cycle — received a structural advantage that had nothing to do with their actual ability. Everyone else was left guessing.
The obvious answer is AI. But existing AI admissions tools have two failures I found unacceptable. First, they upload your most personal data to someone else's servers. Second, they hallucinate — happily inventing a 92% GPA for a student who has 65%, or fabricating awards that don't exist. For a tool whose purpose is to help someone get admitted honestly, that is disqualifying.
Privacy is architectural, not a promise. The application makes no external network calls for any student data. It cannot leak what it never transmits.
The tool may never lie about the student. A dedicated verification system locks the student's real facts and validates every AI output against them before it is ever shown.
The global private tutoring and college prep market is estimated at $30B+. The Ivory Index eliminates the $3k–50k consultant fee entirely. Zero cloud cost means marginal cost per user is near-zero — Ollama runs on the student's own hardware. The business model can be one-time purchase, institutional license, or freemium.
This project demonstrates that local LLM inference + structured profile injection can produce admissions-caliber output without hallucination — using only commodity hardware and open-weight models. No proprietary API, no cloud, no training required.
Live Dashboard · Alex Johnson · 100% Profile · Ollama Online · 15 Models
Feature by feature: the architecture, the specific engineering problems encountered, and the code-level decisions made to solve them. Screenshots show the actual running application with live AI output.
