I'm a Python developer with a background in backend development and data engineering. I enjoy figuring out why things aren't working as expected, especially when the code runs but the data looks wrong.
I'm also a keen runner and tennis enthusiast, so I'm naturally drawn to sports data and the technology behind fitness apps.
RunWx is my personal project comparing race pace and weather across five courses and 61 editions. View the comparison.
It uses Python and dbt on Google Cloud (GCP): Cloud Storage for files, BigQuery for analysis, and Cloud Run for processing jobs and the read-only API behind the website.
Much of the work involves validating messy source data, avoiding duplicate loads and independently checking warehouse results. I also built Airflow workflows and tested the historical pipeline on Cloud Composer.
It's still evolving. The comparisons explore patterns rather than establish cause and effect.
- Prefect · Merged — Added an option to clear deployment schedules while preserving the default behaviour, with CLI and SDK tests.
- Stravalib · Merged — Found and fixed handling that could ignore a Strava API rate limit, while keeping custom limiters compatible.
- Kloppy · Merged — Found and fixed dataset filtering that changed player timelines and minutes played in the original data.
- dlt · Open PR — Found incremental state loss when switching between dictionaries and Arrow/pandas data; proposed a fix to prevent duplicate loads.
- python-garminconnect · Merged — Traced broken HTTP 404 handling to missing response details and fixed empty-result handling for gear requests.
- DataBallPy · Merged — Found and fixed failures when saving NumPy game metadata to JSON, with save/load regression tests.



