Pilot a new task
Find out whether a small model can handle a task beyond our ready-to-use solutions. We agree the model, what a useful result looks like and a testing period, then run the pilot.
AI USE CASE TEST
Your use case.
A defined pilot.
Not sure a small language model can deliver what you need? We agree a specific task, model, testing period and success criteria with you, then test it and show you the results. This is a separate service from a Loxia Grid deployment.
For teams exploring AI that runs inside their organization.
Available worldwide. The test runs on Loxia’s servers using non-sensitive data only.
What do you need?
Evaluate your model
Already have a model in mind? Tell us the model and the task so we can scope an evaluation around your expected outcome.
Discuss fine-tuning
Adapt a model for a specific task. We review your goal and the available training data, then agree the fine-tuning work and evaluation.
Already know your model and only need to run it across your hardware? Talk to Us ↗
From your question to pilot results.
Choose one task
Tell us what you want the model to do and send non-sensitive sample or mock data. We define what a good answer looks like with you.
Agree the model and testing period
We agree a specific model, success criteria and a defined testing period before the pilot starts. We then run the agreed tests using non-sensitive sample or mock data on Loxia’s servers. Fine-tuning is scoped separately.
See the results
Review the answers and where the model falls short. Results apply to the sample data tested. We explain the next steps before any wider rollout.
Questions before you start
What data can we send?
Non-sensitive data only. Mock data is welcome if it reflects your task. Keep private records inside your organization; do not send them for this test.
Where does the test run?
On Loxia’s servers. We choose the model and handle the test remotely for companies worldwide. You do not need to buy hardware or arrange an on-site visit.
What will we receive?
Sample answers, a review of what worked and what did not, and our recommendation for the next step.
Does the test include training the model?
A standard use case test does not include fine-tuning. If you need model training, choose Fine-tuning / Enterprise Package in the form. We agree that scope separately.
What if the model cannot do the job?
You still see the results and learn where it falls short. The test helps you decide what to do next.
Do we have to buy Loxia Grid afterwards?
No. The test is separate. A later Loxia Grid deployment can run AI inside your organization, supporting data sovereignty by keeping your data under your control.
How long does the test take?
A pilot can be completed in 2 to 3 weeks once we have agreed the scope and received the required sample data. Timing depends on the task, model, data readiness and coordination with your team. We confirm the schedule before starting.
What would you like your model to do?
Leave your details and describe your task. We’ll contact you to agree the model, testing period, success criteria and quote. Fine-tuning can be discussed as a separate Enterprise scope.