These are production deployments, not pilots and not letters of intent. Where an outcome comes from a partner’s reported operational data rather than our own measurement, it says so.
A Note on Evidence
In public-safety AI almost nobody can show their largest deployments properly, because the data is confidential and the customers are police and government agencies. The result is a market where every vendor’s claims are effectively unfalsifiable, including ours and pretending otherwise is how buyers get burned.
So we do three things. We attribute operational outcomes to the partner who reported them rather than asserting them ourselves. We publish the architecture and the scale even where an account cannot be named. And we sell the pilot rather than the claim: a deployment on your own feeds with success criteria agreed in writing is the only comparison that survives procurement scrutiny.
The four operational metrics IREX quotes are reported by MetroCCTV, an IREX partner in the UK and Ireland: 1,000+ cameras per operator, search results in under a second across petabytes, a 10x or better response-time reduction and a 100x or better investigation-time reduction.
Evidence Briefs
Operational need. Detect people on tracks, train surfing, dangerous platform-edge crowding and unattended luggage in the visual conditions of underground stations. Deployment evidence. Specialized training datasets were developed with the transport authority and from synthetic 3D environments. Published outcome. The canon records reliable detection in those complex conditions; it does not publish a customer-quotable accuracy figure.
Operational need. Enforce traffic rules and see congestion across a national road network. Deployment evidence. IREX multi-camera traffic analytics run nationwide, including Ulaanbaatar, for red-light, stop-line and lane-direction violations, congestion detection and average-speed calculation between junctions. Published outcome. The canon identifies national production scope, but does not publish a violation-count or accuracy result.
Operational need. Protect power, oil, gas and logistics sites without a wholesale camera replacement. Deployment evidence. The deployment combines perimeter-intrusion detection, fire and smoke detection, restricted-zone monitoring and integration with existing camera and sensor networks. Published outcome. The canon confirms the production use cases; it does not identify the sites or publish outcome metrics.
Production Deployments
The National Police and municipal forces including Lima and Miraflores use IREX as the backbone of public-safety operations, unifying thousands of cameras into a centralized command infrastructure and cutting both response and investigation times. Modules: facial recognition, traffic analytics with two-wheeler detection critical to the region, weapon detection and video search and investigations. A large-scale national rollout runs in partnership with Visium.
IREX detects life-threatening situations in real time across the Underground: people falling onto railway tracks, train surfers, dangerous crowding at platform edges and abandoned luggage. Specialized training datasets were built with the transport authority and by generating synthetic 3D environments, achieving reliable detection in the complex visual conditions of underground stations.
Electrical power stations, oil terminals and pipelines, gas distribution facilities and logistics sites, protected with perimeter intrusion detection, fire and smoke detection and continuous monitoring of restricted zones. Integration with existing camera and sensor networks makes it an upgrade path rather than a replacement program.
Multiple jurisdictions including Oceanside and El Cajon run IREX for traffic congestion and incidents, weapons in public spaces and monitoring for registered offenders and unauthorized loitering near sensitive locations, with real-time suspect matching through integration with local law-enforcement databases. IREX has also been selected as a school-safety vendor in Humble, Texas and is deployed in a Florida pharmacy chain.
Integrated airport security and business intelligence: suspect recognition against watchlists, unattended luggage detection in terminals and gate areas, weapon detection at access points and crowd density monitoring during peak travel periods, in an environment where thousands of passengers transit daily and demand is continuous.
The first country to deploy IREX multi-camera traffic analytics for nationwide enforcement, spanning the country including Ulaanbaatar. Red-light violations, stop-line infractions and lane-direction controls, plus real-time congestion detection and average-speed calculation between junctions.
The full catalog of AI modules deployed for public safety and law enforcement across participating cities, serving as the central intelligence layer for the country’s urban modernization initiative, with real-time alerting and inter-agency collaboration.
The standard security platform for all national stadiums in the Premier League. Banned football fans are identified against watchlists in real time as they attempt to enter matches, with additional modules covering public-safety threats within and around stadium perimeters on match days.
Facial recognition helps local retail chains prevent millions in losses from organized shoplifting, matching individuals against a database of known offenders on entry and alerting security staff in real time. The deployment demonstrates applicability beyond public safety into commercial security.
Often yes and we will arrange it where the customer permits. Many of these accounts are police and government agencies that cannot be named publicly, which is why we also publish architecture, scale and measured outcomes for de-identified deployments.
Because they are MetroCCTV’s reported operational data, not our measurements. An attributed figure is quotable; a self-reported one in this market is not and we would rather hold ourselves to that standard than be caught by it.
Yes and de-identified reference material exists specifically for accounts that cannot be named. Ask and we will send what is cleared for your situation.
Keep Reading
Every deployment above started as one site and a set of criteria agreed in writing. That is still how it works.
