Vision AI for connected fleets.
GPS tells you where a vehicle is. It cannot tell you that the driver is tired, looking at a phone, or closing too fast on the car ahead. Telox Gulf is building an AI safety layer that reads live dashcam video and vehicle telemetry, and flags risk while there is still time to act.
The problem we are solving
Video that nobody watches
Fleets record thousands of camera hours every day. Footage is usually opened only after an accident, to assign blame, never in the seconds before, when it could have prevented one.
Tracking without behaviour
Location and speed data miss the behaviours that actually cause crashes: fatigue, distraction, phone use and tailgating.
AI that does not fit the Gulf
Driver-monitoring models trained on Western data struggle with Gulf conditions: strong glare, local headwear, regional vehicle types and road layouts.
What the platform does
Three parts, one console, on top of the tracking and dashcam estate we already operate.
AI video telematics
Computer vision on road-facing and driver-facing video to detect drowsiness, distraction, phone use, missing seatbelts, forward-collision risk and lane departure.
Telemetry intelligence
Machine learning on GPS, speed, harsh braking and sensor data to score driver and vehicle risk and spot anomalies across the fleet.
Fleet safety console
Detections become live alerts, driver scorecards and coaching, pushed to fleet managers on web and mobile the moment they happen.
How it works
From camera to fleet manager in one continuous pipeline.
- CaptureDashcams and vehicle sensors stream video and telemetry.
- IngestStreams arrive at our platform in real time.
- GPU inferenceVision and ML models analyse every stream.
- AlertRisk events are scored and raised instantly.
- ActManagers see, call and coach from the console.
Inference is designed to run both in the cloud and on in-vehicle edge devices, so detection keeps working where mobile coverage drops.
Built on NVIDIA AI
Telox Gulf is a member of the NVIDIA Inception program. Our roadmap builds on the NVIDIA stack, running NVIDIA GPUs on AWS.
NVIDIA DeepStream
Multi-stream video decoding and inference across thousands of concurrent dashcam feeds.
NVIDIA TAO Toolkit
Fine-tuning driver-monitoring and ADAS models on our own Gulf fleet footage.
NVIDIA Jetson and TensorRT
Optimised in-vehicle inference where connectivity is weak and latency must stay low.
NVIDIA GPUs on AWS
Cloud capacity for model training and for scaling real-time inference as the fleet grows.
Our roadmap
Today, on-device detection comes largely from our camera vendors. Over the next year we are building our own models, trained on our data and tuned for the Gulf.
First quarter
Stand up the GPU training pipeline, label fleet footage into a training set and fine-tune our first driver-monitoring model.
Second quarter
Run real-time detection on a pilot fleet and launch the AI safety tier in the console, with driver scorecards.
Third quarter
Move inference into the vehicle on edge devices and roll the AI tier across our installed base.
Fourth quarter
Add more detection classes and ADAS features and extend across the GCC.
Where it matters most
Transport and logistics
Public transport, delivery and haulage, where driver fatigue and distraction put passengers and cargo at risk.
Oil and gas
Contractor and field vehicles on long, remote routes, where safety rules are strict and incidents are costly.
Government and public sector
Public fleets that need live accountability and dependable evidence.
Join our AI pilot fleet.
We are selecting fleets to run real-time detection first. Tell us about yours.