An engineer who reads.
A short story of how I ended up cleaning the foundations of data instead of painting their dashboards.
Hi, I'm Cristóbal. Data architect, analytics engineer, and AI product builder. I work from Madrid with a short list of clients at a time.
I studied engineering and ended up, almost without noticing, working on the least glamorous business problems: why two systems looking at the same thing don't agree, why a report takes forty seconds, why nobody trusts the number they're looking at.
I spent ten years in large companies — Bacardi among them — translating what business folks wanted to ask into what the data could actually answer. Independent since 2023.
A timeline.
What I believe.
Less is more defensible
A model with twenty well-chosen tables endures; one with two hundred gets abandoned.
Lineage isn't decoration
If a number can't show where it came from, it doesn't belong in an executive report.
Automate the boring, not the delicate
AI is a great hand for what's repetitive. Business judgment is still made by someone with a face.
Talk to whoever uses the data
If I don't talk to the analyst who opens the report on Mondays, I have no idea what I'm building.
What I work with.
Snowflake · BigQuery
When volume matters.
dbt
Standard analytical layer.
Airflow · Dagster
Idempotent DAGs.
Python · SQL
Explore and commit.
Claude · OpenAI
Agents with tests.
pgvector · Weaviate
Retrieval with telemetry.
Metabase · Power BI
Auditable preference.
Git · GitHub
Everything in a repo with CI.