pathocodes/berlin-infrastructure-etl-orchestrator — reverse-engineered prompt

Build me an Airflow project that turns raw Berlin location data into one clean flat table for neighborhood clustering.

I want the pipeline to read Berlin source tables like ATMs, hospitals, parks, emergency stations, and kindergartens from PostGIS, then create neighborhood level features such as ATM counts, average distance to the nearest 5 ATMs, hospital accessibility score, park area, nearest fire station distance, and total kindergarten capacity. Distances should be real meter distances using the right Berlin projection, and the SQL should be safe to rerun by dropping and recreating output tables each time.

Please include the weighted accessibility logic where full access counts as 1.0 and partial or assisted access counts as 0.5. Also handle the Schlachtensee and Zehlendorf missing join issue mentioned in the notes.

Set it up like a practical repo with Airflow DAGs, SQL scripts, examples, and a short README so I can run the workflow and understand the final output table.