Space weather · Machine learning · Scientific software
I build models that predict how space weather affects satellites and the systems we rely on.
I'm Kyle Murphy, a space physicist and independent consultant in Thunder Bay, Ontario. I've spent 20+ years studying how geomagnetic storms reshape near-Earth space, at the University of Alberta and NASA Goddard. Today I turn that expertise into machine-learning models, open-source software, and research programs for teams working on satellite drag, orbit prediction, and space-weather risk.
See my work CV / résumé Get in touchWhat I do
Space weather & atmospheric density
Storm-time thermospheric density and satellite drag for LEO constellations, density derived from satellite orbits, radiation-belt dynamics, and ground-induced electric fields.
Machine learning & statistics
Random forests and gradient boosting, time-series forecasting from solar and geomagnetic drivers, physics-informed feature engineering, storm-aware validation, and ensemble uncertainty.
Scientific software & data pipelines
Turning research code into installable, documented, tested Python packages. Multi-mission data ingestion, vectorization and parallel speed-ups, reproducible notebooks.
Research leadership & communication
Project management, proposal development, mission-concept studies, teaching scientific programming, and building community through an international seminar series.
Featured work
Space weather · Machine learning
MLTDM: Machine-learning thermospheric density model
A random-forest model of storm-time atmospheric neutral density, published in Space Weather (2025) and released as an open, installable Python package with a worked example notebook.
Orbit mechanics · Scientific software
CONTIGO: satellite-derived atmospheric density from orbits
A modular framework that turns spacecraft orbit data into energy dissipation rates and effective atmospheric density, validated against the Orekit flight-dynamics library.
Machine learning · Uncertainty
ml_fw: an ML toolkit for space-weather forecasting
Reusable building blocks for time-series ML in heliophysics: lagged features, tuning, diagnostics, and ARIMA-based perturbed-input ensembles for forecast uncertainty.
Open-source · Data engineering
GMAG: ground magnetometer data and induced electric fields
An open-source package that downloads and loads data from several ground-magnetometer arrays into one pandas interface, now with tools to compute geoelectric fields for space-weather hazard work.
Latest notes
- dev-hub Handback Mode: Running Tasks from a Claude Project
- dev-hub: Mission Control for Claude Across My Repos
- Vectorizing and Parallelizing JB2008
- skops - Machine Learning Model Persistence
Work with me
I work with research groups, agencies, and companies on:
- Density and drag modelling: building, validating, or benchmarking thermospheric density models (empirical, ML, or orbit-derived) for storm conditions.
- Machine learning for space weather: designing forecasting pipelines, choosing features and validation that respect storm physics, and reviewing existing models.
- Research software: turning notebooks and scripts into maintainable packages with tests, docs, and CI.
- Proposals and science writing: proposal development, mission science cases, and peer-reviewed papers.
The quickest way to reach me is email.