Home/Platform/AI ENGINE
AI ENGINE · large language models

The model can become yours.

DBRAINS is independent of the underlying model. It starts from the best models on the market and reaches, if needed, a private model that lives inside your infrastructure. Whichever model you choose, it powers the WAVEs and SYNAPSES in the same way.

Three options

One question: where the data goes.

  1. Standard

    Frontier models

    Claude, GPT, Gemini and others, accessible via API. No infrastructure to manage, continuously updated and covering any task. The model is configurable per type of operation.

    Default
  2. Alternative

    Open source models

    Open source models managed through an inference service compliant with GDPR and the AI Act. A more private environment from the base model on, without fine-tuning.

    Self-hosted
The comparison

Three solutions, criterion by criterion.

The choice of model is independent of execution: what changes is where the data goes, the infrastructure and when each solution makes sense. Here are the three compared directly.

Frontier default
Open source self-hosted
Custom Model optional
Where the data goes
Passed via API to the frontier model providers
Stay in a managed European inference service
Don't leave: they run on the dedicated infrastructure
Infrastructure
None: managed by the providers, always up to date
Managed inference service, compliant with GDPR and the AI Act
Server dedicated to the customer, powered on on-demand
Compliance
Per the providers' terms
GDPR and AI Act by design
Maximum: your own model, no external cloud
Cost
Pay-as-you-go via API, no fixed cost
Pay-as-you-go on managed inference
On-demand infra, zero for idle time + DEGG path
When to choose it
The vast majority of cases: quality and task coverage
You need more privacy from the base model on, without historical data
Years of structured historical data and regulated sectors

All three power the WAVEs and SYNAPSES in the same way: the knowledge doesn't change, only where the model runs.

Custom Model · when it makes sense

When AI can truly become yours.

For companies with years of structured historical data, the conversation shifts from "which model do you use?" to "can I have a model of my own?". The result is a private model that answers in your company's language and logic.

Dedicated server

The model is hosted on a server dedicated exclusively to the customer, which powers on when needed and shuts down when the team isn't working. No fixed cost for idle hours.

It's not a self-service process

Selecting and organising the data requires support. DEGG guides the journey from mapping the document estate to validating the results.

Vertical fine-tuning produces value only with sufficient, quality data. The threshold is significant: years of structured material, not weeks.

Do you have a wealth of historical data?

Let's assess together whether vertical fine-tuning is right for you, from mapping the document estate to activating the model.