Suraag — सुराग़, “the clue” — started with one question from an investigating officer: the footage exists, so why does finding one motorcycle in it still take a week?
Suraag turns CCTV and drone footage into answers. A server analyses footage once — every person, vehicle and number plate is detected, tracked and indexed — and then investigators ask questions in plain English, follow a suspect across cameras, trace a plate, and seal court-ready evidence. Shops and businesses use the same engine live: a small box watches the DVR, uploads only the moments with motion, and WhatsApp says what happened within the minute.
We will not take your raw video. Footage stays on your premises or on the server assigned to you; the cloud stores analysis only, the GPU keeps nothing, and the language model never sees an image. We will not dress a statistical match up as an identification: every cross-camera link, look-alike and face candidate is labelled a lead, carries its score, and waits for a person to accept, reject or mark it inconclusive. And we will not let an AI touch evidence — classical enhancement stays evidence-grade, AI reconstruction is stamped “not evidence”, and every sealed report carries a hash a court can verify.
Two kinds of camera owner. Police and forensic units, who need post-incident investigation that survives cross-examination — case files with FIR and legal authority, chain of custody from the first byte, the Section 63 certificate under the Bharatiya Sakshya Adhiniyam, 2023. And retail, jewellery, factories, societies, hospitals and campuses, who need to know tonight if someone is in the shop after hours — without a control room. One platform, one engine, one set of honesty rules.
Small team, Hyderabad, working in the open with the officers and shop owners who use the product. Every user-visible change ships with a live check on real footage. Every claim on the security page was checked against the code before it was published — and the things that are not built yet are listed there too.