High-redshift galaxies ↗
Selecting and characterizing massive, evolved galaxies at 3 ≤ z ≤ 4.5 with hierarchical Bayesian models across multiwavelength photometric catalogs.
Observational astrophysics · Statistical data science
Inference at
Inference at
cosmic scale.
I'm Abtin Shahidi. I build Bayesian models that pull signal out of massive, incomplete data — from high-redshift galaxy surveys to an identity graph with a billion edges.
0M+
Data points modeled
0B
Identity-graph edges
0K+
Behavioral time series
0M
Galaxies analyzed
0
Peer-reviewed papers
Riverside, California
Independent researcher · formerly UC Riverside
Independent researcher · formerly UC Riverside
Currently
An astrophysicist who spent three years shipping data science in industry — and brought the statistics back to the sky.
I did my Ph.D. in observational extragalactic astrophysics at UC Riverside, using multiwavelength photometric catalogs to find and characterize galaxies in the early universe. Then I spent three years at VideoAmp as a Senior Data Scientist, applying the same Bayesian toolkit to imputation, identity resolution, and behavioral modeling at industrial scale.
After a postdoctoral appointment at UC Riverside, I now work as an independent researcher in Riverside — still chasing structure in noisy, high-dimensional, incomplete data.
Bayesian inference
Hierarchical models
MCMC · Gibbs · Stan
Dimensionality reduction
Point processes
Python · PySpark · SQL
Selected work
Four ways I work with data.
The through-line is the same everywhere: build a model that respects uncertainty, then let it read the structure out of the data.
Data science at VideoAmp ↗
Three years as Senior Data Scientist: Bayesian demographic imputation, a billion-edge identity graph, and point-process models of TV and purchase behavior.
Making structure visible ↗
Embeddings, posteriors, and intensity functions rendered so that high-dimensional, uncertain structure becomes something you can actually read.
Applied machine learning ↗
A full graduate course built from scratch — statistics, Monte Carlo, and the algorithms behind modern ML — with lecture notebooks and datasets.
Publications
Selected papers.
Seven peer-reviewed papers in The Astrophysical Journal and its Letters, on galaxy populations, environment, and statistical methods for photometric catalogs.
2023
The Art of Measuring Physical Parameters in Galaxies: A Critical Assessment of Spectral Energy Distribution Fitting Techniques
2022
Deblending Galaxies with Generative Adversarial Networks
2020
Selection of Massive Evolved Galaxies at 3 ≤ z ≤ 4.5 in the CANDELS FieldsFirst author
2020
Bridging between the Integrated and Resolved Main Sequence of Star Formation
2020
Spectroscopic Confirmation of a Coma Cluster Progenitor at z ∼ 2.2
Visualizations
The data, made legible.
A few pieces from the research and the teaching — each one turning a model or a dataset into something you can see at a glance.
Astrophysics
Star-forming main sequence
A three-component mixture model of SFR vs. stellar mass across redshift.
Manifold learning
Color-space embedding
A high-dimensional galaxy catalog projected into three UMAP dimensions.
Galaxy selection
Color–color selection
Massive evolved galaxies isolated in sSFR and rest-frame color space.