CybrainiaAI projects do not fail on models. They fail on data.
Čeština

Data & Cloud Architect · Prague, Czech Republic

AI projects do not fail on models. They fail on data.

A pilot takes a week to build and a model can be swapped out in an afternoon. What cannot be swapped out in an afternoon is the layer underneath: where a number came from, what exactly it means, and whether finance and marketing agree on it. That layer is what Cybrainia builds — a BigQuery warehouse, a semantic layer holding the metric definitions, and measurement that holds up.

Services →About →Contact →

Services

Four things Cybrainia does

All services

  • AI transformation

    working· 4 years. I build with this regularly and can hold my own in a design review.

    From pilot to production. Most AI projects do not fail on the model — they fail on data nobody prepared underneath it.

  • Data platform

    deep· 14 years. I have designed, shipped and operated this in production, repeatedly.

    A BigQuery warehouse, transformations in dbt or Dataform, ingest through dlt. Layers the client's own team can still change two years later.

  • Semantic layer

    working· 3 years. I build with this regularly and can hold my own in a design review.

    One metric definition for BI tools and for agents — as versioned text, outside a format only one tool can read.

  • Measurement

    deep· 15 years. I have designed, shipped and operated this in production, repeatedly.

    GTM, server-side GTM, GA4, Measurement Protocol and consent. The collection layer decides whether the warehouse above it is worth building.

Track record

What this rests on

Sectors
Retail · Telco · Travel
Technologies
BigQuery · dbt · Dataform · dlt · Terraform · Google Cloud · GTM and server-side GTM · GA4
Experience
Twenty years in IT: measurement and web analytics first, then data engineering, now AI transformation. That order is not an accident — it is why the collection layer sits next to warehouse architecture here.
Where and how
Prague, Czech Republic · Czech and English

No client names and no numbers. The engagements that would belong here are under NDA or are running for a client I have not asked. When a reference is cleared it will appear here — until then an empty space is more honest than an invented percentage.

A question with nothing attached to it is welcome too — whether BigQuery is the right answer for your case, say. Sometimes it is not.Contact