Open-source safety AI.
Alongside our guides, we build and maintain free, open-source tools that run computer-vision safety detection on the CCTV cameras a site already has. Everything is on GitHub.
Built for your site, not a generic playbook.
Every worksite is different, so we don't try to ship one canned setup. You pick the detectors that match what's actually on your site, or train your own in about five minutes if you can't find one. Everything runs at the edge, on hardware you probably already have lying around.
Detection built for industrial sites
Models tuned for hard hats, high-vis vests, exclusion zones, and unsafe acts. Less false positives than generic CV.
Simple to operate
Set policies, generate compliance reports, and track a Safety Score that quantifies ROI and liability. No PhD required.
Hundreds of detectors out of the box
Pick what you need from the marketplace, or bring your own model to HUB.
Train a new detector in five minutes
Skip the labelling. Generative AI fine-tunes a detector for a new object from as few as five example images.
Edge AI for every camera you already own.
Install once and any camera becomes smart. Hundreds of detection algorithms run in real time, alerts fire when something goes wrong, and a customisable policy engine keeps everything aligned with your safety rules. Analytics roll up into reports that prove compliance.
A complete safety AI deployment, in a box.
A self-contained kit with a HUB, 4G internet, two cameras, screen, UPS power back-up, and an industrial socket. Everything sits in an IP66 weatherproof enclosure. Built for sites where CCTV and networking are a challenge. Runs on generator power or solar, pole-mountable.
See it in action
Ask any question of any camera, in plain English.
Sentinel applies generative AI and LLMs to live CCTV feeds. It recognises activity and records it as searchable text, so you can query feeds in natural language and get answers, ChatGPT-style.
Try SentinelThe boring stuff that actually prevents injuries.
Pull up any incident log and you'll see the same handful of patterns. A hard hat that came off. Someone standing way too close to a forklift. A guy who walked into an exclusion zone because the cone was knocked over. We watch for those on the camera feeds you already have, ping whoever needs to know in real time, and the events get rolled up into a weekly report your EHS team won't ignore.
Explore case studies