Academic background

I got my undergrad degree in physics from Carleton College. I worked as a researcher in astronomy both at Carleton and through the NSF REU program at Lowell Observatory. After college, I went to Boulder to do my Ph.D. at the University of Colorado. I was a postdoc at the University of Minnesota from 2011-2016, and was briefly a Visiting Scholar at the University of Kentucky.

In 2016, I left academia for a fellowship at Insight Data Science. I've worked as an Applied Scientist at Amazon in Seattle since 2016.

Data Science

My work as a research and applied scientist in industry includes predictive pricing models, ranking features, high-cardinality ML classification at large-scale, causal inference, and large language models (LLMs).

Astrophysics

Galaxy morphology

Galaxies are the baryonic building blocks of the Universe, including our own Milky Way. The morphologies of galaxies, ranging from smooth ellipticals to grand design spirals, are snapshots of the composition and kinematics of the gas, dust, and stars within the galaxy. These probe the physics that govern the formation and evolution of galaxies. I used data from the Galaxy Zoo citizen science project as probes of morphologies in very large surveys, including SDSS, COSMOS, and UKIDSS. I led the data processing for two major phases of the project: Galaxy Zoo 2 and Galaxy Zoo: Hubble. My scientific publications probe the relationship of small-scale disk structure to broader galaxy properties, such as environment and star formation rate.

I was the first head of data pipeline development for Radio Galaxy Zoo, which leverages volunteers to match radio lobes from active black holes to their host galaxies. I'm keenly interested in citizen science as a discipline through the Zooniverse, and have developed tools for volunteers to help them better engage with the data and the scientific community.

Blazar environments

Blazars are a class of active galaxy where a relativistic jet is viewed almost directly into our line-of-sight. This allows direct measurement of many high-energy processes, but also obscures details about the host galaxy. I used SDSS images to study the local environment of different types of blazars; their clustering shows that blazars do exist in overdense environments, meaning that nearby interactions likely play a role in triggering the jet action. However, sub-types of blazars such as FSRQs and BL Lac objects don't show any statistical difference in the distribution of their galactic neighbors, pointing instead to internal parameters such as the accretion rate.

OH megamasers

My PhD research focused on properties of OH masers: these are are radio-wavelength lines powered by stimulated emission of radiation. OH megamasers (OHMs) are found in the nuclear regions of (ultra)luminous infrared galaxies and are caused by pumping of giant molecular clouds by infrared photons. If the gas densities are low with a favorable geometry, this causes a population inversion in the OH molecule that results in a maser. Overlapping clouds along a particular line of sight produce powerful emission lines that can be 10,000 times brighter than the total radiation from our own Sun.

The relationship of the OHM to its host galaxy, however, is not well understood — while there are many infrared-bright galaxies in the universe, only a very small fraction of them host an OHM. As a result, the question of what distinguishes a non-masing ULIRG from a masing ULIRG is still open — what are the specific triggers that cause an OHM? To address this, we took mid-infrared spectroscopic observations of a large sample of OHMs and non-masing galaxies using the IRS on the Spitzer Space Telescope. Mid-infrared data is key for several reasons: 1), mid-infrared photons are capable of penetrating deep into the dusty nuclear regions of these galaxies where the masers are formed; 2), there are a wealth of features in the mid-IR that are powerful diagnostics of the physical conditions; and 3), mid- and far-infrared photons are thought to dominate the pumping process of the OHM, according to current models.

Using the IRS, we assembled a large sample of mid-infrared data on more than half of the known OHMs. The data shows significant differences between the masing and non-masing populations, particularly in the dust optical depth and the slope of their 20–30 μm continuum. Theoretical pumping models agree with our distribution of measured dust temperatures, but severely underpredict the necessary dust optical depth. We suggested that this is due to a preferred dust geometry; comparing the IRS spectra to radiative transfer models show that OHMs are well-fit with a smooth, enveloping shell of dust. Non-masing galaxies, on the other hand, show shallower silicate absorption depths and ratios consistent with a clumpy torus often associated with AGN.

I also worked extensively with the Green Bank Telescope to do a large survey of infrared galaxies at high redshift to search for new OHMs, and used the Very Large Array to survey Andromeda for OH masers in an attempt to measure the galaxy's proper motion.

Compact symmetric objects

Compact symmetric objects (CSOs) are radio-loud galaxies characterized by jet/hotspot activity on either side of a central engine. The spatial extent of these radio structures are extremely small compared to classical radio galaxies (less than a kiloparsec) and have advancing proper motions that can be measured with VLBI imaging. This results in kinematic ages for CSOs that are extraordinarily young — on the order of hundreds to thousands of years. CSOs thus likely represent the very earliest stages of radio-loud galaxies that may later evolve into an AGN.

I investigated the mid-infrared properties of the known low-redshift CSOs using infrared spectra from the Spitzer Space Telescope. CSOs are a remarkably diverse class of galaxies in the mid-infrared, with characteristics ranging from a pure starburst to a continuum-dominated AGN. Most interestingly, although energy balance arguments suggest that the jets could be powered by Bondi accretion onto a central black hole, we found no direct evidence (such as high-ionization atomic emission) in any CSO for accreting gas. This raises the possibility that young radio galaxies and CSOs may be powered by more exotic mechanisms, such as the tapping of black hole spin.

Pulsars

As an undergraduate, I worked on two projects involving pulsars (rotating neutron stars). In 2004, I worked with Prof. Joel Weisberg with observations at Parkes Observatory in Australia of the pulsar B1641-45. Observing both the on and off spectrum of the pulsar beam in the OH line showed direct evidence of stimulated emission in the interstellar medium. This also placed constraints on the spatial scale of OH clouds, suggesting that molecular clouds are clumpier than those of neutral hydrogen.

On another observing run at Parkes, I helped to characterize the properties of a sample of southern pulsars at high frequencies (8.4 GHz) with Dr. Simon Johnston. The polarization fraction of most pulsars at this frequency turned out to be low, with the exception of young, energetic pulsars.

Spatial scales in irregular galaxies

As an REU student at Lowell Observatory, I worked with Dr. Deidre Hunter on analyzing Fourier transform power spectra in irregular galaxies. Power spectra characterize the spatial scales of structures in an image; for a galaxy, this gives clues to the types of star and galaxy formation taking place. We showed that most irregular galaxies have power spectrum slopes consistent with Kolmogorov turbulence. The slopes are also similar to those seen in dwarf and giant spiral galaxies, suggesting that the microscopic processes are independent of galaxy size or the presence of spiral density waves.