- 🤓 Scientist + developer interested in climate/weather AI
- 🐍 I specialize in Python, ML, and designing complex data science applications and pipelines
- 🌨️ MSc in atmospheric sciences, and BSc in physics + astrophysics
- 🚀 Interested in scaling ML pipelines (multi-GPU/multi-node with Lightning AI, PyTorch, MLFlow, Comet ML)
- 🌎 Currently deeply involved with designing CanESM infrastructure with Integrated Modelling System Infrastructure (IMSI)
- 🤖 Tinkering with GANs, stable diffusion, ConvNets (sorry, not all open source!)
- 👨💻 Canadian Centre for Climate Modelling and Analysis, Environment and Climate Change Canada 🇨🇦
- 👷 Climate Data and Analysis Section, Environment and Climate Change Canada 🇨🇦
- 🏛️ University of Victoria Climate Lab 🇨🇦
- 🌎 Pacific Climate Impacts Consoritum @pacificclimate 🇨🇦
- 🔭 Herzberg Astronomy and Astrophysics, National Research Council of Canada 🇨🇦
- 🌃 Dark Cosmology Centre, Neils Bohr Institute, Denmark 🇩🇰
- Operational climate/weather AI that forecasts actionable information (accurate, fast, high-resolution, and on relevant timescales)
- Better means for storing and processing multidimensional data
- Purpose built ML models
I have been writing software for approximately 10 years. In that time I've noticed some themes that I'm trying to learn from as part of my continual growth as a scientific software developer.
- Great science follows from great infrastructure, not necessarily the other way around.
- Reproducibility is a cornerstone of good science. Clarity and simplicity in programming work together towards reproducible science. Too often do scientists exempt themselves from these practices because science is hard. The fact that science is hard only strengthens the need for clarity and simplicty in programming.
- "Premature optimization is the root of all evil" - Donald Knuth
- Annau et al. (2023) - Algorithmic hallucinations of near-surface winds: Statistical downscaling with generative adversarial networks to convection-permitting scales Artificial Intelligence for the Earth Systems, 2(4), e230015. DOI: 10.1175/AIES-D-23-0015.1
- Climpyrical, scientific software computing National Building Code of Canada design values
- Design Value Explorer Web application for visualizing and downloading design value fields and tables.
- R Package ‘ClimDown’
-
Herwig et al. (2018) - Cyberhubs: Virtual research environments for astronomy The Astrophysical Journal Supplement Series, 236(1), 2. DOI: 10.3847/1538-4365/aabfe7
-
Monthly Notices of the Royal Astronomical Society, 503(1), 176–199. Thomas et al. (2019) - Dwarfs or giants? stellar metallicities and distances from ugrizg multiband photometry DOI: 10.1093/mnras/stab499
-
The Astrophysical Journal, 886(1), 10. Thomas et al. (2020) - VizieR Online Data Catalog: Dwarfs or giants? Stellar metallicities & distances (Thomas+, 2019) VizieR Online Data Catalog, J–ApJ. DOI: 10.3847/1538-4357/ab4a7f
-
Higgs et al. (2021) - Solo dwarfs II: the stellar structure of isolated Local Group dwarf galaxies DOI: https://doi.org/10.1093/mnras/stab002
-
Christensen et al. (2017) - Solving the conundrum of intervening strong Mg II absorbers towards gamma-ray bursts and quasars Astronomy & Astrophysics, 608, A84. DOI: 10.1051/0004-6361/201731340
- ECMWF 2022 Machine Learning Workshop
- CMOS 2022, Computational Methods Machine Learning and Model Development: Extreme Super‑Resolution and Downscaling of Wind Fields at Convection‑Permitting Scales
- 6th Spatial Statistics Generative adversarial networks for super‑resolving near‑surface wind patterns
- CMOS 2024 Leveraging AI for Enganced High-Resolution Regional Climate Modelling: ClimatExML: Designing AI Software for the Computational Demands of High‑Resolution Climate Models