Veebase

Safety inspections across Middle East worksites are still documented with phone photos and handwritten Arabic notes, typed up hours later — slow, inconsistent, and easy to miss a missing guardrail or a worker without PPE. Baseer (بصير) by Veebase is an Arabic-first field inspection copilot. An inspector uploads a photo or short video, or runs a live camera; Baseer segments the scene, reasons over it with Meta's models, and produces a structured bilingual report in about 30 seconds — every finding linked to its exact frame, with severity and a corrective action. Every finding opens a case (Open → Assigned → Verification suggested → Closed) that closes only after re-inspection evidence AND human approval. An evidence chat answers questions grounded strictly in findings and frames, and an append-only audit log records every action. Built for the Meta Global AI Developer Hackathon, Startup track, Build for Your World. The demo was filmed on the vision path, and the fixed 24-scene benchmark results are published in the repo README.