# One team from design to production

The engineers who map your problem are the ones who design, build, deploy and run the system. There are no hand-offs between consultants, architects and developers.

## Every skill a system needs, in one team

- **AI engineering**: Agents, retrieval over documents, MCP servers and evaluation.
- **Data engineering**: Pipelines and data transformation.
- **Cloud**: Infrastructure as code, deployment pipelines and identity.
- **Enterprise applications**: SAP, NetSuite, Dynamics 365 and Odoo.
- **Governance**: Access control and audit trails.
- **Product and UX**: Dashboards and apps people use every day.

## Four habits on every project

- **Always measure**: Every AI-related change passes an evaluation before it ships. On the catalogue assistant every change is tested on 50 real questions, and the social-data pipeline has twelve release gates.
- **AI only where it is needed**: Ordinary code where the answer must be exact, models where language or images are involved. In the pipeline, rules match genres and keywords and a model reads posts only for artist names; the KPIs of the distributor are defined once, in SQL.
- **Governance by design**: Access rules, policies and audit from the first sprint. The catalogue assistant reads the ERP as the signed-in user, and the KPI connector acts as the user who asks and can only read.
- **AI-native delivery**: Our engineers direct coding agents, and a separate pass checks everything the agents write.

## AI evaluation before every release

The evaluation is how we show that a system works. The same four steps run on every project, with tools adapted to the problem.

- **Agree what good means**: We agree with you, before we build, which answers count as right and how often they must be right.
- **Build the test set from real cases**: Questions, documents and records taken from your own work.
- **Measure every change against a baseline**: Each change is compared with the version before it, so a regression shows before release.
- **Release on the result**: A change ships only when it meets the bar.

## Platforms we build on

- **Microsoft Azure**: The production base of every system we deliver.
- **AWS**: Cloud services for clients who run on Amazon Web Services.
- **Claude, by Anthropic**: The platform our AI training runs on.
- **Microsoft Dynamics 365**: Business Central, one of the ERPs our systems connect to.
- **Odoo**: The ERP Velais itself runs on, and one our systems connect to.
- **SAP, NetSuite and ElevenLabs**: SAP Business One and NetSuite, which our systems read from, and ElevenLabs for voice AI.

## What we do not do

Our work is AI engineering and AI training, so printers, networks and computers stay with your IT team. After a course, Claude runs on your company's own licence, because we do not resell licences. The training ends with a working skill for each person and no certificate.

## Talk to us

Book a call with us.