Nara Abroad

Arab students and their school counsellors lose days on study-abroad applications. They copy grades from certificate photos, work out whether a Tawjihi score is enough, pick courses, fill in forms and chase missing documents. Many families pay an agent USD 300 to 500 just for advice. Nara Agent runs on Meta's models. It does the counsellor's paperwork, and the counsellor approves every field. It is Arabic first, with an English toggle. How it works: 1. Interview in Arabic. The student answers by voice or text. The agent builds a profile: grades, stream, goals, budget, preferred countries and English level. 2. Document photos. The student photographs their Tawjihi certificate and English test report. SAM 3.1 finds and crops the document. Muse Spark reads every field and where it sits on the page. 3. The agent works on its own. In at most 12 steps, it converts the Tawjihi grade, searches sample programmes, checks entry requirements, fills the form for the best match and lists what is missing. A live trace shows each step. 4. Counsellor review. Each field shows its value, a confidence level and its source: a zoomed crop of the document with the value highlighted, or a quote from the interview. Fields with no source are marked "needs counsellor". The counsellor approves or edits each field, then the pack. 5. Class mode. A counsellor runs a whole class: up to 30 students, 3 at a time. One table shows each student's best programme, entry route and missing items. What we built during the hackathon: We started too early: about one to two hours before kickoff (26 Oct, 16:00 UTC), before we had checked the official start time. That work, a first offline version (26 Oct, 13:59 to 14:58 UTC), never called a Meta model. The first real calls came after kickoff, on 26 Oct at 21:10 UTC. Built during the hackathon: document reading with evidence crops, fixes from the first live runs, the English toggle, class mode, and the README and demo script. Every commit, with UTC and ET times, is in the README. Sample data for the demo: the universities, programmes and admission rules are fictional, the certificate images are synthetic, and no real student data is used anywhere.