Aaditya GC
Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

Sacho Sathi (साँचो साथी — "a true friend") is a Nepali-first AI evidence checker that helps people verify suspicious forwarded messages before believing or sharing them. In Nepal, messages about scholarships, government benefits, exam updates, and job offers spread rapidly through Facebook, Viber, and family messaging groups. Some are genuine — many are outdated notices recirculated as current, claims stretched beyond their real scope, or announcements with no official source at all. Checking them takes time, confidence, and the ability to navigate formal official websites, which many recipients don't have. So the message gets forwarded anyway. Sacho Sathi solves this with a single flow: upload a screenshot or paste the message, and the system extracts the checkable claims, searches for evidence prioritizing official Nepalese sources (ministries, universities, the Public Service Commission, Nepal Rastra Bank), compares each claim against what it retrieved, and produces an Evidence Receipt in simple Nepali or English. Each receipt gives a claim-by-claim verdict — Supported, Contradicted, or Insufficient Evidence — with real source links, mismatch explanations (expired Bikram Sambat dates, changed amounts, widened eligibility), a downloadable receipt image, and a respectful Nepali reply the user can send back to the group without embarrassing the sender. It never invents a source: every citation is validated in code, and when evidence is weak, it honestly says "insufficient evidence" instead of guessing. In evaluation on 46 labeled claims plus screenshot tests, it correctly handled supported, contradicted, outdated, and unverifiable forwards — including catching last year's scholarship notice through BS-to-AD calendar conversion, something no generic chatbot does.