Skip to content
Kwun Chen · Data Science
Kwun Chen

Data Science + Statistics at UC Berkeley. I build pipelines that make noisy data usable.

adsb_pipeline.py
Three F-15 Eagles flying in formation against a pale sky

Military Aircraft Activity Pipeline

Complete

Live ADS-B telemetry turned into clean, analysis-ready flight features

  • Python
  • Polars
  • Pydantic
  • DuckDB
  • Prefect

The idea

ADS-B transponders broadcast an aircraft’s position, altitude and speed every few seconds. The raw feed is noisy and inconsistent. This pipeline turns it into a table you can actually analyze for military aircraft activity.

How it works

  • Ingest. Each run pulls real-time telemetry from a live global ADS-B feed and processes 500+ aircraft per run.
  • Validate. Pydantic schemas check every record on ingest, so bad rows get caught before they reach the warehouse.
  • Transform. Polars computes flight-activity features from the raw telemetry: climb rate, acceleration, heading change, and a regional classification.
  • Warehouse. The results are loaded into DuckDB.
  • Orchestrate. Prefect runs the validation, transformation and warehousing stages as one flow.
← All projects