Merlin · AI safety supervision
A second set of eyes that never blinks.
Merlin watches the cameras your site already has and catches safety problems while there is still time to act: a worker with no hard hat, mask or vest, a spill, a blocked exit, a fire. An AI reads the frames and names what it sees. The right person gets notified. You keep a record you can hand to an inspector.
Today
Safety rests on supervisors who cannot watch every camera, every worker, all the time. Violations go unseen, near misses go unrecorded, and there is no easy way to prove what happened.
With Merlin
Passive feeds become an active, auditable safety record. Issues are flagged as they happen, routed to the right person, and kept as a defensible history. "We think we are compliant" becomes evidence.
These are industry-wide figures, not our results. They are why a second set of eyes on every camera, every shift, is worth having.
recordable workplace injuries and illnesses were logged by US private employers in 2024.
BLS, Survey of Occupational Injuries and Illnesses, 2024US workers died on the job in 2024, about one every 104 minutes.
BLS, Census of Fatal Occupational Injuries, 2024a year goes to the ten leading causes of serious workplace injury, more than a billion dollars a week. Same-level falls rank second.
Liberty Mutual Workplace Safety Index, 2025 (2022 data)fall protection has been OSHA's most-cited violation, with eye and face PPE in the top ten.
OSHA Top 10 Cited Standards, FY2024Watch
An appliance on your network reads the camera feeds you already have.
On-siteDetect
An AI reads the frames and names what it sees: missing gear, hazards.
CloudApply your rules
A vest on the line, none needed in the break room. Your zones decide.
On-siteNotify the right person
Email, in app alert or webhook, with approval gates when you want them.
On-siteSign off
A supervisor logs what they actually saw. Now it is evidence.
On-siteFire, person down and blocked exits can never be dismissed automatically. If your floor needs something else watched, tell us and we add it to your site's categories.
When something is wrong, the right person hears about it.
A finding becomes an alert: emailed to a supervisor, posted in app, or sent to a webhook. Here is what two of them look like.
No hard hat
Lathe operator, Bay 3. The cloud AI called it, your rules said it counts here, and it went to the floor supervisor for sign-off.
Spill detected
Coolant on the floor near CNC 2. Read from the frame by the cloud AI and logged, with maintenance and the shift lead notified.
Every call you close makes the next shift quieter.
Alarm fatigue is what kills safety systems: a tool that cries wolf gets muted. Merlin does the opposite. The longer it runs on your floor, the less it bothers you about things that are not a problem here.
When a supervisor closes a finding as not a concern and writes why, that reason becomes part of your site policy. It is a closed loop, so Merlin gets more trustworthy the longer you run it.
Flagged
A welder in a face shield trips the "no mask" rule.
Corrected
A supervisor closes it as not a concern and writes the reason.
Learned
That reason becomes site policy. The same call stops coming back.
Uses the cameras you have
Nothing to rip out. One appliance on the local network reads the feeds you already have.
Your records stay on site
The AI that reads the frames runs in the cloud, configured for zero data retention. Your findings database and audit trail live in the building.
Evidence you can defend
Every finding carries two signals: the AI that called it and a person who signed off on what they actually saw. A record, not a black box score.
Merlin is a safety observation aid that supports supervision. It flags and observes. It does not replace machine guarding, supervision or lockout/tagout, and every metric it reports is descriptive.