Senior Data Engineer

Context

Axiom Analytics provided white-label marketing analytics to 80+ SME clients in Spain and Portugal. I joined as the second data engineer to productionise a prototype attribution engine that the data science team had built in Jupyter notebooks.

What I did

Refactored the attribution notebooks into a modular Python package, deployed as an Airflow DAG on Cloud Composer. Each client's data was partitioned in BigQuery by client_id with column-level access controls. Built a lightweight dbt layer for standardised marketing metrics (CAC, ROAS, LTV) shared across all client schemas.

Introduced automated SLA monitoring: if a client's pipeline missed its 06:00 delivery window, PagerDuty fired before the client noticed. Reduced SLA breaches from ~8/month to 0 over a 6-month window.

Outcomes

• —Attribution model latency: 4 hours → 22 minutes.
• —SLA breaches: 8/month → 0 over the final 6 months.
• —Codebase went from 3,000 lines of notebook cells to a tested, documented Python package with 87% test coverage.