Careers
What we do

Artificial Intelligence & Process Automation

Raise productivity and unlock new business opportunities with intelligent algorithms and automated workflows.

A digital brain connected to a robotic arm and a data panel
What we do

Where AI comes in

Three families of use cases. We always start from a sample of your own data, never from a demo built to impress.

01

Machine Learning & Predictive Analytics

Algorithms that learn from your data to forecast scenarios and optimise processes.

  • Predictive maintenance on plant and machinery
  • Demand and consumption forecasting
  • Quality control on process data
02

Robotic Process Automation (RPA)

Automation of complex, repetitive workflows across different software systems.

  • Data extraction from documents and e-mail
  • Reconciliation and data entry across systems
  • Automated approval workflows
03

Computer Vision & NLP

Solutions for image analysis and natural language processing.

  • Visual inspection of parts on the line
  • Reading delivery notes and supplier invoices
  • Document assistants over your own archives
How we work

From a data sample to a model in production

Four phases with a declared exit point: if the accuracy is not there, we stop before you spend.

01 · week 1

Sample and feasibility

A hundred photos of parts or a month of documents are enough to say whether the case is tractable and with what error margin.

This assessment is free.

02 · weeks 2-5

Proof on your data

We train a first model on your data and measure the accuracy reached.

If it is not enough, we stop here and say so.

03 · weeks 6-14

Integration into the process

The model enters the real workflow, with a human check wherever an error has consequences.

04 · ongoing

Monitoring and retraining

Performance is measured over time: when the data shifts, the model has to be retrained. That is part of the contract, not an extra.

Technology

Technology

Mature, documented technology: the code has to stay maintainable by people who did not write it.

Machine learning
PyTorch
Language and documents
LLM API RAG
Numerical computing
MATLAB Fortran
Automation
RPA
FAQ

The questions we are asked most

How much data do we need to start?

Less than people expect: a hundred images for a visual check, or a month of documents for data extraction, are enough for a first feasibility assessment.

Do our data end up in an external service?

Only if you decide so. Where the data is sensitive we run the models on European cloud or on a server at your site, and the boundary is put in writing before we start. Your data do not train third-party models.

What happens when the model gets it wrong?

The error margin is stated and measured before release. In critical cases the model proposes and a person confirms: the automation does not decide on its own.

How long does the initial assessment take?

One week, and it is free. At the end you know whether the case is tractable, at what accuracy and at what order of magnitude of cost.