Origin
David Bramante built the original Factory Guardian prototype solo in less than nine hours for the Antler × Zapdos Labs American Industrial Revolution Hackathon, where it earned a first-place tie. The prototype watched a camera feed, analyzed visible factory conditions, logged events, and sent alerts. The commercial opportunity is broader and more practical: configure each camera around the questions an operator already asks while reviewing footage.
Working Prototype
The original build established the end-to-end interaction. The next phase replaces the temporary hackathon inference path with production camera connections, facility-specific rules, measured accuracy, and operating safeguards.
Operational Questions
The system should not begin with a vague promise to understand an entire factory. Each camera gets a small set of observable rules tied to a real management decision.
Count people inside a defined work zone and record when staffing falls below the facility's expected level for longer than an agreed threshold.
Measure how long a station is unattended or visibly inactive so managers can investigate recurring workflow gaps without scrubbing through an entire shift.
Flag when a service counter or other customer-facing zone remains uncovered during scheduled operating hours.
Watch for specific, observable conditions—such as required PPE in a marked zone—and send the relevant clip to a person for review.
Example Rule
During scheduled production hours, notify management when fewer than three people are visible in a defined department for longer than five consecutive minutes, then report the total understaffed time for the shift.
That rule is specific enough to test against recorded footage, tune for a real camera angle, measure false positives, and decide whether the resulting information is operationally useful.
Human Review
Camera analytics become dangerous when a system turns uncertain visual signals into confident judgments about people. Factory Guardian should preserve a clear boundary between observation and decision.
Design-Partner Path
Start with the department where better visibility would create the clearest operational value.
Translate what a manager looks for into precise conditions, zones, time thresholds, and review requirements.
Run a baseline period, review false positives, and establish whether the camera angle can answer the question reliably.
Add cameras and workflows only when the first deployment produces useful, trusted information.
FAQ
No. Factory Guardian is an early CodexDeus-owned venture with a working hackathon prototype and public demo. Production camera integrations, facility-specific rules, validation, security controls, and operating workflows are part of the commercial pilot path.
The preferred approach is zone-level and process-level observation without facial recognition. A facility can measure coverage, unattended time, and defined events without turning the system into an identity tracker.
It can measure defined visible conditions and flag clips for review. It should not infer intent, declare that a person is productive, or accuse somebody of theft from ambiguous video.
The current design-partner focus is an operator with an existing facility, accessible cameras, a recurring observable problem, and a willingness to define and validate one narrow workflow before expanding.
Facility Operators
The best first pilot starts with one camera, one recurring question, and a result your team can verify. Factory Guardian is developed by CodexDeus in Austin, Texas.