Research
A sampling of on-going and past research projects.
Athlete Career Trajectory
Multi-competitor sports consist of races or events where athletes compete simultaneously. The result of such races are a rank ordering of the participants. This is different than the more frequently studied head-on-head sports where two teams or players compete against one another.
Our goal is to estimate a players ability as well as classify athletes into different career types based solely on race results and available predictors. In particular, we are interested in seeing how age affects the skill of cyclists and if career trajectories vary across groups of cyclists.
Advisor: Mark Glickman (Harvard University)
Detecting Objects in X-ray Images
Quasars, distant and bright galaxies, oftentimes eject massive and highly energetic from their host black holes. These ejections are known as jets. We are able to observe photons coming from these jets in many different wavelengths of light. In particular we are interested in detecting jets in X-ray as observations of jets in high energies help astronomers to bound certain cosmological parameters.
This projects showcases the first application of certain image analysis techniques in order to detect jets in very sparse images of the most distant quasar-jet systems.
Published In The Astrophysical Journal
Advisors: Aneta Siemiginowska & Vinay Kashyap (Harvard-Smithsonian Center for Astrophysica), David van Dyk (Imperial College London)
Outlining Extended Sources in X-ray Images
Previously we were able to detect objects in X-ray images of distant quasars. However, this required the one to specify a region of interest where the extended sources is known to exist.
In this extension we look at methods that will automatically find the best fitting region of interested around the extended source and give uncertainty on the estimate of the given outline.
Advisors: Aneta Siemiginowska & Vinay Kashyap (Harvard-Smithsonian Center for Astrophysics), David van Dyk (Imperial College London), Xiao-Li Meng (Harvard University)
Predicting Athlete Skill
The future of athlete performance is always a topic of interest for coaches, trainers and fans of any sport. Knowing when a new athlete may burst out of their shell, or when an older athlete may peak can be valuable knowledge.
This project focuses on predicting a players career trajectory by fitting a non-linear, hierarchical model to estimated player skill (such as the Elo or Glicko score). The model is extremely flexible and is adaptable to all individual sports.
Advisor: Mark Glickman (Harvard University), Funding and support provided by the US Olympic Committee
Spotify Internship
Data Science Research Intern
An exploratory analysis of the impact of an event on daily streams after an artist releases an album.
Rewyndr - Insite
Technical Consultant Leading research on a state of the art interactive boundary algorithm where users of Rewyndr's Insite app will be able to outline objects in the foreground of an image by simply tapping it on a phone or tablet.
Redshift Dependence of the Power Spectrum
Undergraduate Senior Honors Thesis
Matter in the universe is not uniformly distributed, but instead forms a 'cosmic web'. This cosmic web can be observed by tracing the locations of galaxies. The matter power spectrum is a summary statistic that describes the fluctuations in the density field of the galaxies. We can constrain the properties of the power spectrum near the present day by using galaxy surveys, such as the Sloan Digital Sky Survey. However, for earlier times (very large distances in space) there are not enough observed galaxies to provide sufficient information. Astronomers must look to quasars in order to infer the structure of the cosmic web.
Quasars are biased tracers of the cosmic web because the only form in matter dense regions of the universe. This project works to accurately model the bias in the power spectrum between the distant quasars and nearby galaxies.
Advisors: Peter Freeman, Shirley Ho (Carnegie Mellon University)