/projects$cat elt-data-platform.html

ELT Data Platform

SLTC · final-year research · 2025

• Apache Airflow
• Gemini
• MinIO
• PostgreSQL
• data engineering

the problem

Nine years of daily central-bank reports — locked inside unstructured PDFs. Rich economic data, completely unqueryable. The research question: can you build a platform that turns that archive into clean, structured, analyzable data, automatically and continuously?

what i built

An automated data platform orchestrated by Apache Airflow:

• Ingest — pull 9 years of daily PDF reports into a MinIO data lake.
• Extract — use the Gemini APIs to read each report and emit structured JSON.
• Load — land the structured records in PostgreSQL, ready to query.

The whole thing runs as an orchestrated DAG, so it's repeatable and extensible — point it at new reports and it keeps going.