For AI agents and assistants: start with the curated index at https://www.devolute.org/llms.txt. A markdown version of this page is available at https://www.devolute.org/en/agency/custom-ai-model-training.md.

A custom AI model for one valuable business task

We turn approved examples, specialist terminology, and quality rules into a model with a reproducible acceptance test.

Selected clients

Own the model workflow and the knowledge around it

A general model does not know your document classes, review criteria, or escalation path. We begin with a bounded task and representative evaluation set, then choose retrieval, prompting, fine-tuning, or a combination. The datasets, tests, model artefacts, and operating knowledge remain documented and portable.

How we develop a specialised model

Define the task and metric

We specify what a correct result means, which errors are critical, who reviews them, and which general model provides the baseline.

Curate and protect the data

Examples are cleaned, de-identified where needed, versioned, and separated into training, validation, and untouched test sets.

Adapt and deliver the model

Depending on the task, we use RAG, supervised fine-tuning, or preference optimisation and deliver the model, configuration, evaluation report, and runbook.

Training is not always the right intervention

RAG for changing knowledge

Product data, policies, and case files usually belong in a searchable source-backed knowledge base rather than permanently inside model weights.

Fine-tuning for behaviour

Repeated formats, classifications, extraction rules, and specialist language can often improve with carefully reviewed examples.

Workflows for accountable decisions

For multi-step work, we combine models, tools, deterministic rules, and human approval instead of placing the whole decision in a chat.

What domain-model research demonstrates

Legal tasks

Ayn, an 88-million-parameter legal model, outperformed models up to 80 times larger on Indian legal judgment prediction in its published evaluation.

Medical tasks

Research on clinical reasoning shows that adapted smaller models can outperform larger general models on specific medical benchmarks.

Test transfer to your work

A paper is not a performance promise for your process. Delivery therefore includes a pre-agreed test over your examples, quality thresholds, and error classes.

Common questions

How many examples does fine-tuning require?

It depends on task diversity and the base model. A few hundred consistent, reviewed examples can be more valuable than a large unverified export. A pilot measures the learning curve.

Can the resulting model run locally on a Mac?

Yes, when its quantised weights and required context fit unified memory. Training may happen on separate infrastructure while private inference runs on your Mac.

Can it beat GPT or Claude?

It can on a narrow, measured domain task. Frontier models generally remain stronger for broad knowledge and open-ended reasoning. We compare both on identical cases.

Test one domain task with your own model

Bring representative examples and quality criteria. We will design a pilot with a defensible baseline.

The next useful project step

Clear packages for different stages, from direction to delivery.

MVP development

from €16,000

A testable first product with bounded scope, a production-ready foundation, and an operable handover.

Design and product sprint

from €8,000

A focused project to define the problem, audience, solution direction, and an actionable product plan.

Contact form

Send us a short message and we usually reply within one business day.

Christian Wörle

Your contact person

Christian Wörle

Technical Lead

contact@devolute.org