Mike Smith
AstroAI Fellow at UniverseTBD
Cambridge, Massachusetts, United States
Summary
👤
Senior🎓
Top SchoolMike Smith is an AstroAI Fellow and co-founder with 9 years of experience applying self-supervised, unsupervised, and generative models to unusual data modalities, especially satellite and astronomical imagery. Based in Cambridge, MA, he leads AI product development at Aspia Space—authoring a patented generative ClearSky algorithm for SAR-to-visible translation—and now develops open ML models for astrophysics at the Center for Astrophysics | Harvard & Smithsonian. His PhD work produced high-impact generative tools (GANs and diffusion models) for synthetic astronomical surveys and ultra-fast galaxy profile extraction, with outputs featured by NASA and published in leading journals. Comfortable moving between research and product, he has spun research into commercial offerings like EarthPT, the first temporal Earth-observation foundation model, and downstream models for biomass, land use, and yield prediction. Adept at turning niche scientific problems into scalable ML systems, he combines deep academic rigor with entrepreneurial drive and a knack for making “weird” modalities tractable.
9 years of coding experience
3 years of employment as a software developer
Doctor of Philosophy - PhD, Astronomy and Astrophysics, Doctor of Philosophy - PhD, Astronomy and Astrophysics at University of HertfordshireMPhys, Physics, First, MPhys, Physics, First at University of LeedsGCSEs and A Levels, GCSEs and A Levels at Dame Alice Owen's School
Github Skills (24)
denoising10
diffusion-probabilistic10
foundation-models10
generative-model10
astronomy9
machine-learning9
transformer8
deep-learning8
multimodal8
artificial-intelligence8
spect7
galaxy6
diffusion-models6
earth-observation6
pytorch6
Programming languages (6)
Github contributions (5)
Smith42/astroPT
Feb 2024 - Jul 2026
Transformer based foundation model for astronomy
Contributions:11 releases, 5 reviews, 56 PRs in 2 years 5 months
astronomytransformerlarge-observation-modelfoundation-models
Smith42/astroddpm
Oct 2021 - Jan 2022
A denoising diffusion probabilistic model synthesises galaxies that are qualitatively and physically indistinguishable from the real thing.
Contributions:1 release, 28 commits, 11 pushes in 2 months
denoisingdiffusion-probabilisticmachine-learningastronomygalaxies
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.