Forge by SkillsBase brings custom computer vision into an accessible workflow for inspection teams. Users can label images, train models and build a process for field users without writing code.
Computer vision has enormous potential in field operations, but many organisations have found it hard to apply. General-purpose models are not trained on their equipment or their faults. Building a custom model has traditionally required specialist machine-learning skills, carefully prepared datasets and significant development time. And even a capable model is of limited value until it is part of the process engineers follow on site. Forge is designed to address each of those barriers.
Starting with the inspection requirement
The platform starts with the organisation’s own inspection requirement. Teams define the equipment and visible conditions they want to check, then prepare examples that reflect the work. Their operational knowledge helps determine what the model should recognise and what needs a person’s judgement.
That starting point keeps the project grounded. Rather than asking what AI could do in general, the team asks a specific question about its own work, such as whether a component is present or whether an installation detail meets the required standard, and builds from there.
The Forge workflow
Forge brings the stages of building an inspection model into one place:
- Label: mark up example images to show the model what it should recognise.
- Train: build a custom model from those labelled examples.
- Build: create an inspection workflow around the model for field users.
- Deploy: make the workflow available on mobile devices, including on-device use where supported.
Forge connects that model with the steps an engineer follows during an inspection. The aim is to make the check usable within the task, with relevant guidance and a clear way to review the result. Supported models can also run on the device where a reliable connection is unavailable.
Because these stages sit together, teams can move from an idea to a working trial without handing the project between separate tools or suppliers. Results from the field can then inform the next round of labelling and training.
More than a model
This end-to-end approach is the focus of the platform introduction. The model is one part of the process; deployment, the inspection instructions and the handling of uncertain cases also need attention before a team expands its use.
Forge grew out of SkillsBase’s experience with the Virtual Buddy, developed for Openreach to give engineers immediate feedback on their work. That project showed the value of feedback at the point of work, and the importance of designing around the engineer’s task rather than the technology. Forge takes those lessons and makes them available to any organisation with its own inspection requirements.
Meet Forge at Connected Britain
Forge was shortlisted for the AI Innovation Award at Connected Britain 2026. Visitors can meet SkillsBase at Stand K13 to explore the platform and discuss a specific inspection task. Bring representative examples and the questions your quality team needs answered so a demonstration can focus on your requirements.
Connected Britain takes place on 9 to 10 September at ExCeL London. If you cannot attend, the SkillsBase team is happy to arrange a demonstration at another time.
To learn more, visit the Forge website or contact the SkillsBase team.