Seeing the Human Layer: Helin Brings Edge Vision AI to Industrial Operations on Azure

A joint solution from Helin and Microsoft, built on Azure IoT Hub, Azure Device Registry, and Microsoft Fabric. Available on Azure Marketplace.

Industrial operations are rich in machine data and blind to the people working alongside the machines. On a drill floor, the top drive, the iron roughneck, and every powered system report their status to the driller, while the roughnecks moving in between them go unrecorded. That blind spot is where safety incidents and lost productivity hide, and closing it is where the next gains in industrial operations will come from.

TL;DR

• Helin and Microsoft are launching a joint solution that brings edge vision-AI to industrial operations on Azure.
• Built on Azure Machine Learning, Azure IoT Hub, Azure Device Registry, and Microsoft Fabric.
• First application: RedZone, hazardous-zone monitoring live on offshore drilling rigs, alerting in roughly 150 milliseconds with no cloud round-trip.
• Available now on Azure Marketplace, and co-sell ready for Microsoft account teams.

A platform, not a point solution

Helin did not build a single safety product. It built the platform that safety and operations applications run on, and made it repeatable on Azure. Underneath every deployment is one loop: train a model in the cloud, deploy it to the edge, deliver inference where the work happens, and feed new edge data back to sharpen the next model.

The stack:

• Helin edge inference and Data Collector: run vision-AI inference on-site and connect into the industrial control systems already on the asset, so a detection carries operational context. A person is not just “here”; a person is here, during this operation, with this equipment active.
• Azure Machine Learning: trains and versions the vision models before they are deployed to the fleet.
• Azure IoT Hub and Azure Device Registry: the device plane Microsoft is building next, managing and governing each edge box as a first-class Azure resource across the fleet. Helin has run production workloads against this integration since its public preview.
• Microsoft Fabric: Eventhouse ingests the movement and telemetry time-series streaming up from the edge in real time; Fabric turns that time-series into the operational model, the analytics, and the reporting.

“The availability of a built-in certificate manager is a great upgrade in keeping the IoT space more secure.”

Martijn Handels, CTO, Helin, on Helin's early work with Azure IoT Hub's device registry preview

The pattern: process at the edge, and stream only the signal that matters to the cloud. Inference runs on the box, on-site. Nothing depends on the cloud at runtime, so the edge application keeps running and acting through network outages, and raw video never leaves the asset. The platform turns video into movement data locally, and only that data syncs to Azure. For an offshore rig on a constrained satellite (VSAT) link, that is not a convenience. It is the only way the system can work.

Proven in the hardest environment

The first application built on the platform is RedZone, a hazardous-zone monitoring solution deployed on offshore drilling rigs. It detects when a person enters a danger zone around active heavy equipment and alerts in roughly 150 milliseconds, with detection and alert happening on the box and no round-trip to the cloud. It needs no wearables or tags, and it logs timestamped evidence for audits automatically.

Offshore, safety-critical, surrounded by moving steel, on connectivity you cannot rely on: if the pattern holds here, it holds almost anywhere. Helin customers in offshore drilling already use it to watch operations live, support crews in real time, and keep drilling safely while avoiding downtime.

Noble implemented Helin’s Remote CCTV Manager to provide a live video stream of 20+ rigs to authorized personnel at Noble’s office locations. Optimized video streams are stored in the secure Azure cloud tenant and full-resolution recordings are stored temporarily at the drilling rig to minimize the bandwidth required and data transmission costs. Read the full case study here.

RedZone is one application. A different customer, with a different hazard, a different zone, or a different rule about who can be where and when, can build their own the same way, on the same platform and the same Azure foundation.

From an alert to a model of the operation

The alert was the beginning, not the product. Once the platform is continuously detecting people and equipment, it is no longer collecting alerts. It is collecting a continuous record of the human layer, the one no system had before.

What turns that record into insight is a data model that relates equipment to operations to people. A detection stops being a set of coordinates and becomes “this person, in this role, during this phase of the job, near this machine.” The data now describes work, not pixels.

That is where it pays off. In periods where traditional analytics were blind, holding process output but no situational awareness, the movement record is the only thing that can account for what actually happened. Post-job analysis becomes possible: reconstruct a finished job hour by hour, chart where people and equipment really spent their time, and surface patterns no machine log would ever show, including recurring bottlenecks, wasted movement, and hazards no one had named. All of it runs on cameras already installed for safety. The same sensors that keep people out of danger now show how the whole operation moves.

Get started

For Microsoft account teams, Helin is a concrete, co-sell-ready solution for customers asking about edge AI, worker safety, or operational efficiency. Reach the Helin team here, or for operators ready to start, visit the Azure Marketplace listing.

For operators ready to build a real-world model of their operation, Helin is available now on Azure Marketplace, visit the Azure Marketplace listing.

Ready to bring the human layer into your Azure data estate? Connect with the Helin team, or find Helin on Azure Marketplace today.