Straight answers, before you have to ask.

Yes. Send us images of the defect that matters to you, and we train a model on your product and report what we find — including when the honest answer is that your current imaging cannot support it. This happens before any commitment and at no cost.

No. xis.dev is a no-code environment built for quality and process engineers, so your team can annotate, train and validate models without any programming or data science background. If you would rather not do it in-house, we can collect the data and build the model for you.

Yes. Some customers do everything themselves in xis.dev; others ask us to capture the image data on their line and build the model for them. Both options are available and you can move between them later, because the platform is the same either way.

Yes. A pilot runs on one line, live and alongside your existing process, so you can compare results directly before committing to further lines. Most customers start with a single inspection point and expand once it has proven itself.

Three components. xis.dev is where your team builds and trains models. xis.rt runs the trained model on an edge device at the line. xis.dashboard consolidates results from every xis.rt instance into a single view across lines, plants and countries.

xis.dev builds the model, xis.rt runs it, and xis.dashboard analyses what it produces. xis.dev is used by your engineering team during setup and whenever the product changes. xis.rt runs continuously at each inspection point, and xis.dashboard is where quality and operations management review performance across sites.

Yes. A single xis.rt instance handles several product variants and several defect types simultaneously — different products running on the same line, each with its own checks, and multiple inspection types applied to one unit. Every check reports its own result against that unit.

Yes. Each xis.rt instance is a complete inspection system on its own, making decisions, signalling the PLC and recording results without any connection to other components. xis.dashboard is added when you need a consolidated view across multiple lines or sites.

Three components per inspection point: an industrial camera, controlled lighting, and an industrial PC with an NVIDIA GPU. Lighting matters as much as the camera — most vision systems that fail on a live line fail on illumination, so we specify it as part of the design.

In most cases, yes. xis.ai retrofits onto existing industrial cameras, lighting and industrial PCs, using standard interfaces rather than proprietary hardware, so equipment already on your line is usually reusable. We confirm compatibility with your specific setup during the feasibility assessment, before anything is purchased.

We help you select and specify the right industrial camera, lighting and industrial PC for your inspection, then integrate them into the line — so you are not working out the optics and illumination on your own.

xis.rt integrates directly with PLC and MES systems over OPC-UA and MQTT. It signals reject, alarm or line halt in real time and passes per-unit results into your MES for traceability. Both are standard industrial protocols, so no custom middleware is required.

Yes. xis.ai is designed to be retrofitted to lines already running, using your existing conveyor, cameras and control system. Each inspection point is added independently, so a line can be equipped without redesigning the rest of the process.

xis.rt runs state-of-the-art AI models at 150 frames per second. Achievable line speed depends on your product, camera and the checks being run, so we confirm the figure for your specific line during the trial rather than quoting one number for every application.

Your team retrains the model in xis.dev. New examples are annotated, and a new version is validated and deployed to the line without raising a ticket or waiting for a service visit. This matters most where recipes, seasons or packaging change regularly.

Whichever you prefer. xis.dev is built so your own engineers retrain models in-house whenever the product or specification changes, which means no waiting on us for routine changes. If you would rather we handled retraining as a service, that option is available too.

Every decision is recorded with its image and measured value, so a false reject or a missed defect can be reviewed directly rather than argued about. Your team tunes thresholds in xis.dev, and those same records provide the evidence for what to adjust.

On your own infrastructure. Inference happens on the industrial PC at the line, and where a dashboard consolidates multiple lines or sites, it is deployed on your servers. No image, result or metadata passes through infrastructure operated by xis.ai.

No. xis.rt operates fully offline at the line, with no cloud service and no internet dependency. A network connection is needed only if you choose to consolidate results into a central xis.dashboard — and that connection is to your own servers.

Yes. xis.dashboard consolidates results, images and metadata from every xis.rt instance regardless of location, giving group quality and operations management comparable data across lines, plants and countries. The dashboard runs on your own servers, so the data stays within your infrastructure.

No. xis.ai operates no cloud service for inspection data. Models run locally, results are stored locally, and consolidation across sites happens on infrastructure you control. We have no access to your production images or inspection records unless you choose to share them with us.

xis.ai does. We help you select the right camera, lighting and industrial PC for your inspection, integrate them into your production line, and build the first AI models, including collecting the image data. Your team then operates the platform, with us available for further models whenever you want.

Handover covers a full platform manual, training videos and working sessions with our engineers, so your team can operate xis.dev independently. Beyond that, we remain available to build additional models or collect further image data as a service, whenever you would rather not do it in-house.

Yes, and most customers do. Each inspection point runs its own xis.rt independently, so lines are added one at a time without redesigning the deployment. xis.dashboard is introduced when you want a single view across several lines, plants or countries.