Get to know us · New Delhi · Mumbai

The intelligence layer above your existing cameras.

Xorth is built by OcelAI Private Limited, a computer-vision company based in New Delhi and Mumbai.

Book a Demo → Our research
Safety Security Operations 1 backbone · 24 models

Xorth is a software developed by OcelAI Pvt. Ltd. the computer-vision research company behind the platform.

24
Detection models
< 7 days
Typical time to first cameras live
RTSP
Connects to the CCTV you already own
Our mission

Safety intelligence built for the sites we actually have

Built by
OcelAI Private Limited
Based in
New Delhi · Mumbai
Deploys
On-premise · private cloud · hosted
Data
DPDP Act, 2023 — you remain Data Fiduciary

Indian sites are not short of cameras. They are short of anyone watching them. Footage gets reviewed after an incident, which makes CCTV a forensic tool rather than a preventive one. We build the software layer that closes that gap — no rip-and-replace, no six-month deployment, and no pretence that a camera pointed at the wrong thing can be fixed in software.

Why Xorth exists

Safety performance is still reported almost entirely through lagging indicators — the injuries that already happened. The leading indicators that predict them are hard to collect because they depend on somebody observing and recording an unsafe act. Xorth makes that collection continuous and automatic, turning passive CCTV into a source of leading-indicator data.

Research & innovation

Computer vision that ships to site.

Xorth analyses your camera feeds continuously and surfaces the events that need a decision. Frame rates, active models and detection thresholds are tuned per camera during commissioning, and reviewed with your team, so the alert stream stays credible instead of becoming background noise.

Models are refined against real site footage, not benchmark datasets, because a helmet at forty metres in Delhi haze is a different problem from a helmet in a research paper.

See how it works →
<Seconds
Detection to alert, not overnight batch
24/7
Continuous coverage
3
Deployment modes: on-premise, private cloud, hosted
1 → 24
One backbone model, 24 site-specific detection models

Safety

  • PPE compliance
  • Work at height & fall arrest
  • Falls & prolonged immobility
  • Crane & lifting proximity
  • Exclusion-zone breach
  • Hot work, fire & smoke
  • Material stacking

Security

  • Perimeter & after-hours intrusion
  • Restricted-area access
  • Loitering
  • ANPR
  • Enrolled-person recognition
  • Material & plant protection

Operations

  • Zone activity
  • Headcount & entity count
  • Covered-load verification
  • Waste handling
  • Crowding
  • Visitor logging
How we work

Three principles behind every product decision.

01

Hardware-agnostic

Works with existing IP CCTV over RTSP through your NVR, from all major camera brands. Analogue systems connect through an encoder. Where a camera's angle or lighting cannot support a model, we say so during the survey rather than after the invoice.

02

Deploy in days, tune in weeks

From signed contract to first cameras live in about a week; a fully tuned site in two to four weeks. Inference needs a GPU server per site — yours, your private cloud, or one we size for you. We publish the tuning period rather than hide it.

03

Privacy by design, not by policy

Xorth runs where your policy requires. In on-premise and private-cloud deployments video stays inside your environment, and only events, metadata and short evidence clips reach the dashboard. You are the Data Fiduciary under the DPDP Act, 2023.

The team

Meet the people building Xorth

Our engineering, AI and field-operations teams work directly with the safety officers, site engineers and security leads who use the platform.

01

Founders

A computer-vision founder and a civil-engineering co-founder — the models and the sites they run on are understood first-hand.

02

Engineering & AI

Building the computer-vision models and inference pipeline that power every Xorth deployment.

03

Field Operations

Site surveys, commissioning, camera health monitoring and support for deployed installations.

04

Customer Success

Working with site managers, EHS leads and security heads to keep detection tuned and the reporting useful as the site changes.

Frequently Asked Questions

How Xorth connects, what it detects, and where your video is processed.
General
How does Xorth work with our existing cameras?
Expand answer
Xorth connects to your existing IP CCTV over RTSP through your NVR or VMS, and to any ONVIF-conformant camera; analogue estates connect through an encoder. There is no camera replacement and no new cabling. In practice this covers all major camera brands — Hikvision, Dahua, CP Plus, Hanwha, Axis and Bosch among them. As a rule of thumb a model needs a 2 MP or better stream, enough light for the zone, and a view in which the subject fills enough of the frame to be resolved — but we confirm that per camera position during the survey rather than promising it in general. Angle, distance and lighting decide which models a position can support, and we tell you which ones it cannot.
How long does deployment take?
Expand answer
First cameras are typically live within about a week of network access. A fully tuned site — zones drawn, detection thresholds set, false positives reviewed with your safety team — usually takes two to four weeks depending on camera count. Tuning is not optional: an alert stream nobody trusts gets ignored, and an ignored alert stream is worse than no system at all.
Where is our video processed and stored?
Expand answer
Wherever your policy requires. Inference can run on a GPU server inside your own network, in a private cloud tenancy you control, or on Xorth-hosted infrastructure — the mode is fixed in your order form, not changed later. In on-premise and private-cloud deployments the video stays inside your environment and only events, metadata and short evidence clips reach the dashboard. Where you choose Xorth-hosted, that infrastructure runs in Mumbai, India, with the region confirmed in your order form. A full audit trail is maintained in every mode. Under the Digital Personal Data Protection Act, 2023 you remain the Data Fiduciary; Xorth acts as a Data Processor on your documented instructions.
What can Xorth actually detect?
Expand answer
Twenty-four detection models across three domains. Safety covers PPE, work at height and fall arrest, falls and prolonged immobility, crane and lifting proximity, exclusion-zone breach, hot work, fire and smoke, and material stacking. Security covers perimeter and after-hours intrusion, restricted-area access, loitering, ANPR, enrolled-person recognition and material and plant protection. Operations covers zone activity, headcount and entity count, covered-load verification, waste handling, crowding and visitor logging. Ask us for the full model list — which of them a site runs is a configuration decision, not a licensing one.
Can Xorth integrate with our existing security and EHS systems?
Expand answer
Yes. Xorth is an additional layer over your existing operation, not a replacement for it. Events are delivered by email, SMS, in-app, open REST API and signed webhooks, so they can be pushed into your incident register, EHS management system, access-control platform or ticketing tool. Deeper connectors are scoped with your systems integrator.
What happens when something is detected?
Expand answer
The assigned responder receives an alert carrying the camera, the zone, a timestamped evidence clip and the model's confidence score. Every alert is retained with its review state, so you can show who saw it, when, and what corrective action was raised. Detection is probabilistic and always requires human judgement — Xorth surfaces the event and holds the record; your team decides what it means.
Xorth