TransitOps Berlin

TransitOps Berlin is a learning project about two things at once: building a small data platform around Berlin public transport, and learning how to work with an AI coding agent without handing it the steering wheel.

Overview

The planned system samples VBB static GTFS and GTFS-Realtime data for Berlin U-Bahn, S-Bahn, and tram services. The data path uses cron for real collection, Parquet and DuckDB for storage and analysis, dbt for models and tests, and an isolated Airflow demo for synthetic orchestration.

Current state

Stage 1 is complete. The production pipeline and application are still being built, so this page does not pretend there is a finished control room to try.

The repository keeps the project decisions, prompts, skills, acceptance criteria, checks, and deviations as curated evidence. The goal is to make the work inspectable, not to publish a raw agent transcript.