vinimeurer/teste-tecnico2 — reverse-engineered prompt

I need you to build this take home project for me as a complete data pipeline for a logistics company.

Use Python with PySpark as the main processing engine, and make it run in Docker with a single command, ideally docker compose up. Read the raw files from the data folder in all the formats that are there, clean the bad records, handle duplicates, nulls, broken references between entities, invalid GPS points, impossible speeds, and standardize dates and text fields.

The important part is enriching the GPS tracking with the geofences, figuring out when a truck is inside a fence or on the road, and detecting entry and exit events during each trip. Then create a solid enriched trips dataset plus the business metrics the README asks for, like trips by month and status, average trip time by route, average speed, delay rate, top drivers, fleet utilization, and average stopped time by geofence type.

Please organize the outputs in raw, staging, and trusted or analytics layers, keep it idempotent, configurable, and update the README with clear run instructions and architecture notes. Look up current docs online if you need to.