John F. Wu, Phd Email & Phone Number

Baltimore, Md, Us

Baltimore, Maryland, United States

I am currently a tenure-track Assistant Astronomer and the AI Scientist in the Data Science Mission Office. between Jan 2022 and Aug 2024, I was an Instrument Scientist in the Roman Telescope Branch of the Instruments Division.

Baltimore, Maryland, United States

I am a postdoctoral member of the ISM*@ST research group, and I focus on understanding how galaxies grow and evolve by using massive imaging data sets and sophisticated machine learning methods. I am collaborating with various other research groups in order to study various galaxy populations, including dwarf galaxy satellites, rare local analogs of very distant galaxies, and other gas-rich systems. Using novel deep learning methods, I have identified the largest ever sample of faint nearby galaxies, with which we can investigate cosmic substructure in unprecedented detail.

Baltimore, Maryland, United States

I created a novel way to learn about a galaxy's cold gas content based on just an image and achieved state-of-the-art results by combining convolutional neural networks for regression and pattern recognition. I also worked on three ways to investigate and interpret the neural networks:- use Gradient-weighted Class Activation Maps to identify salient galaxy features in the images- analyze the encoded representations using dimensionality reduction and data visualization tools- probe correlations between variables (e.g., galaxy properties) by studying out-of-distribution neural network performanceThis work will appear in the Astrophysical Journal; see https://arxiv.org/abs/2001.00018 for a preprint, or the blog post for a light introduction to visualizing the trained models.

Greater New York City Area

My thesis research included small independent projects to large international collaborations. I worked on multi-wavelength observations from all kinds of telescopes, and analyzed data using models ranging from simple Bayesian regression to deep convolutional neural networks. - Trained and tested machine learning models to recognize galaxy properties from astronomical imaging- Cleaned and analyzed data from heterogeneous sources, including the Hubble, Herschel, and ALMA telescopes- Performed statistical analyses in order to constrain physical models of galaxy growth and evolution- Developed and tested software pipelines for the calibration and imaging of the new MeerKAT telescope array (LADUMA survey)See my blog post on using deep learning to better understand elemental abundances in other galaxies below.

Greater Pittsburgh Area

Analyzed the stellar properties of a million cluster galaxies by writing a custom Python-based pipeline.

Implemented and tested new facial recognition software based on PCA.

Doctor Of Philosophy - Phd, Astrophysics

Bachelor Of Science - Bs, Physics/Astrophysics