Computer vision for sports spectators
We developed a mobile-optimised computer-vision model used by spectators at sports events (see Players to Crowd under our work).
A company-specific AI model for a clearly defined task
We build AI solutions that use your company’s data and are scoped to a precise task. Model size, architecture and runtime environment are determined by the task.
Who the service is for
The service suits organisations with a valuable use case, suitable data or the ability to collect it, and the capacity to invest in experimentation, training and deployment.
- Organisations with a clearly defined AI use case
- Organisations with usable data or the capability to gather it
- Larger organisations and public bodies through a limited work package, subcontracting or membership of a wider project team
The typical situation
The company has usable data, repeated inspection work or a detection problem that is too heavy to do by hand — but the technical implementation is unclear.
Typical tasks
- Object recognition in images or video
- Classification
- Anomaly detection
- Quality or measurement-data assessment
- Prioritising cases that require human review
- Bounded text, classification and prediction tasks
- Integration into a mobile app, device, web service or server environment
Delivery process
A step-by-step process where risk is kept small in the experiment phase.
1
Use-case definition
We determine what problem the model should solve and what success means.
2
Data assessment
We examine the quality and sufficiency of the available data.
3
Limited experiment
We run a bounded experiment to determine whether implementation is realistic and what it would require.
4
Integration
We connect the finished model into the service, application or process.
5
Monitoring and controlled updates
We monitor model behaviour and deploy updates in a controlled way.
Practical examples
Image generation
SekoKuva® is an image-generation service that runs an actual image-generation model in our own runtime environment (see our work).
Price and principle
Price and schedule are assessed after clarifying the use case and the data situation. The first discussion is free.
Model size, architecture, training resources and runtime environment are determined by the task, data and accuracy target.
Book an AI-model feasibility assessment →Limitations
- We do not promise absolute locality or perfect security without project-specific definition.
- Chatbots are not excluded, but the service is not positioned primarily as chatbot development.
Frequently asked questions
- Do we need ready data before starting?
- Readily available data is not always required, but having data or the ability to collect it is a practical precondition. We assess your situation in the first phase.
- How large a model is needed?
- It depends on the task, the data and the accuracy you need. A small, tightly scoped model is often enough; sometimes a larger one is required. The right size is established in the assessment, and we promise nothing in advance.
- Can the model run in our own environment?
- Yes. The solution can be hosted in a managed cloud, your own infrastructure or an on-premises environment. The options are compared per project.
- What happens after the feasibility assessment?
- You receive a report assessing feasibility, required resources and a realistic accuracy target. The investment decision is your company’s to make — we bind you to nothing before that.
- Does the service also suit larger organisations?
- Yes. Larger organisations and public-sector projects are delivered as a limited work package, through subcontracting or as part of a wider project team.
Does your company have a technical objective or problem, but no clear solution yet?
Describe the situation in your own words. We will determine whether we can help and how the work should proceed.