Luis Carbonell Gironés

Led the migration from dbt Cloud to self-hosted dbt Core on Astronomer Airflow with Astronomer Cosmos, eliminating dbt Cloud licence costs, and designed a tag-based scheduling system that resolves model dependencies automatically.

Designed deployment pipelines that wait for in-progress runs and exclude the models being deployed from later runs until the deployment finishes, so releases never collide with scheduled production runs.

Own the CI/CD platform for the Data and ML teams: built the shared standards tooling and reusable GitHub Actions workflows used across 20+ repositories, with security and quality scanning (Semgrep, Bandit, Trivy, TruffleHog, Hadolint, Kubeconform, Kubescape, Kubesec) and Renovate dependency automation.

Deployed self-hosted GitHub Actions runners on EKS with the Actions Runner Controller (ARC), building a custom ECR base image with pre-installed dependencies that cut CI startup time.

Implemented CI/CD workflows that provision an ephemeral Astronomer Airflow deployment per pull request, reducing infrastructure costs and improving developer experience.

Built reusable ingestion tooling that onboards source tables as dbt models and runs Presidio PII detection, enforcing GDPR data classification at the ingestion layer.

Designed a greenfield dbt Core project on Trino, enforcing layering, naming, data quality, and materialisation standards through CI checks and pre-commit hooks.

Run the pipelines behind finance reporting, Account Manager revenue dashboards, ML features, and Braze marketing data, and review the team's pull requests daily.

Owned the CI/CD pipelines deploying Airflow DAGs to Amazon MWAA via S3, with dependency management and shared coding standards for reliable production orchestration.

Contributed to the platform migration from Snowflake toward an Apache Iceberg lakehouse with Starburst Galaxy/Trino as the compute layer, applying cost-aware modelling and query optimisation.

Migrated legacy SQL pipelines to dbt using incremental processing and source partitioning, reducing pipeline execution costs.

Removed duplicate datasets in the feature layer by converting them to views, reducing storage costs and improving maintainability.

Led the end-to-end re-platforming of the ELT stack from Airbyte/dbt Cloud to Dagster, dltHub, and SQLMesh, reducing pipeline execution costs by 38%.

Built a production CDC pipeline from scratch with Debezium and Kafka, capturing changes from PlanetScale (Vitess) and delivering them to BigQuery through a custom dltHub integration, moving ingestion from batch to near real-time.

Introduced SQLMesh audits plus dbt contracts and tests that catch schema and type mismatches at build time, stopping corrupted data from propagating downstream.

Redesigned core data models as a star schema, eliminating duplicate datasets and reducing storage costs by ~13%, while keeping downstream consumers working through view abstraction.

Established CI/CD pipelines with code quality checks, automated testing, and deployment across the data platform, and documented sources and models with full lineage and metadata.

Designed and built ELT pipelines with Airbyte and dbt, adding automated data quality checks to improve pipeline reliability.

Introduced Git-based version control and code review for dbt models and ELT pipelines on GitHub.

Turned statistical analysis of internal metrics into new fraud-detection rules, and built Metabase and Sigma dashboards for business teams.

Supported the team in gathering customer-satisfaction insights and developing internal dashboards for visualisation.

Conducted data preprocessing, knowledge mining, and statistical analyses, communicating insights to stakeholders.