Data platforms that cost less, run faster and answer in plain language.

I'm Victor Urquiola, a Senior AI Data Engineer with 8+ years building lakehouses on Snowflake and AWS, and shipping production AI on top of them.

Decision. Ingestion writes to S3 Iceberg and the warehouse reads it, instead of loading everything into the warehouse.

Trade-off. Of three options, this took 85% of the savings at near-zero complexity. Leaving the warehouse saved more but added scripts and write bugs.

Featured work

How I integrate the tools, and for what kind of solution.

AI-Powered Knowledge Platform

For teams whose knowledge is scattered across meetings, tickets and repos.

Anyone can ask in plain language and get answers grounded in that knowledge, without interrupting an engineer.

Model chosen by A/B test: 4× faster at P95, tool success 100% against 50%.

  • RAG
  • Vector DB
  • Hybrid search
  • ReAct agent
  • LLM observability

Governed Lakehouse on AWS

For organizations that need shared data under strict security rules.

One governed place to model and query data, with orchestration and data quality checks automated.

Clean datasets, aligned with security and compliance requirements.

  • S3
  • Glue Catalog
  • Athena
  • dbt
  • Airflow
  • Lake Formation

Multi-channel Analytics Platform

For businesses selling across several channels.

Unifies stock and sales from every channel, including sources with no API that only send spreadsheets by email.

Data freshness from 24 hours to about 60 seconds.

  • Snowflake Streams
  • Dagster
  • dbt
  • Fivetran
  • n8n

Issue-to-PR Runner

Personal project. For engineering teams that want agents on real work.

Carries a refined issue to a reviewed pull request, without taking the person out of the merge.

First milestone shipped in September 2026. Every attempt is traced and scored.

  • Agents
  • Claude Code
  • Git worktrees
  • Langfuse

Personal Finance Operations

Personal project, in use since July 2026.

Replaces reading card statements line by line and rebuilding a spreadsheet every month.

From hours to about 6 minutes a month, my own measurement.

  • Python
  • FastAPI
  • HTMX
  • Postgres
  • DuckDB

Open-source Lakehouse on Kubernetes

Personal lab. For teams that want a lakehouse without vendor lock-in.

The same Bronze, Silver and Gold layers as a cloud warehouse, built only from open-source parts on Kubernetes.

Deployed from Git with Argo CD and Helm, on a cluster built with Terraform.

  • Trino
  • Iceberg
  • Nessie
  • MinIO
  • dbt + Cosmos
  • Kubernetes
  • Argo CD
  • Terraform

Also delivered: event-driven pipelines on Azure Databricks

A form submission triggers a Databricks job that processes the data and notifies the team in Teams. Deployed with Asset Bundles and CI/CD in Azure DevOps.

  • Databricks
  • Delta Lake
  • Logic Apps
  • Azure DevOps

How I work

Three principles behind how I solve problems.

  • Kaizen

    Continuous improvement and constant learning.

    I tackle a problem in small iterations, measure each one and keep what works: every iteration is a chance to do better than before.

  • Kodawari

    Attention to detail and excellence in execution.

    Before calling something done, I chase the edge cases, the failure modes and the small inconsistencies. The small things define the quality of the whole, which is why I break every problem into smaller, decoupled parts.

  • Ownership

    Thinking like an owner of the product and the outcome.

    I follow a problem beyond my part until it is solved, and I answer for the result, not only for the technical delivery.

Stack

Data platform architecture, pipelines, lakehouse, analytics engineering and AI automation.

Platforms

  • Snowflake
  • Databricks
  • Postgres
  • DuckDB
  • Iceberg
  • Delta Lake
  • Trino

Transform and orchestrate

  • dbt
  • Airflow
  • Dagster
  • Python

Ingest

  • Airbyte
  • Fivetran
  • n8n

Cloud and infrastructure

  • AWS
  • Azure
  • Google Cloud
  • Docker
  • Kubernetes
  • Argo CD
  • Terraform
  • GitHub Actions

AI

  • Snowflake Cortex
  • Vector DB
  • Anthropic
  • OpenAI
  • Langfuse
  • Claude Code

Serve

  • Power BI
  • Looker
  • Superset
  • Streamlit
  • FastAPI

About

I started in 2018 writing ETL jobs. Today I own data and AI platforms end to end: architecture, build, deployment and operations.

I care about what a system costs, how it fails and whether someone can trust its numbers.

Spanish, Portuguese and English.

Certifications on LinkedIn
  1. 2026 to nowVictor Urquiola Labs, Senior AI Data Engineer (Independent)
  2. 2024 to nowMutt Data, Senior Data Engineer
  3. 2021 to 2024Cobli, Data Engineer
  4. 2019 to 2021RockingData, Junior Data Engineer
  5. 2018 to 2019Logyt, Junior ETL Developer