Research

Faint Kuiper Belt Objects

Variational Autoencoders

Probabilistic Cataloguing

We implement a Bayesian method for producing point source catalogues. Rather than obtaining a single catalogue, a probability distribution of catalogues that is consistent with the data is inferred. This probabilistic catalogue better captures the inherent uncertainties in point source identification. We use this method to constrain unresolved point source populations and better de-blend stars in crowded fields.

Statistics for the Turbulent ISM

Characterizing the Galactic Centre GeV Excess

As a part of event reconstruction, the Fermi-LAT Collaboration estimates the accuracy of the direction reconstruction of each event, as certain events suffer from inefficiencies and inactive regions in the LAT. By creating new analysis classes of Fermi events by imposing a cut on direction reconstruction quality, the tails of the point spread function can be reduced at the expense of effective area. This method is applied in an analysis of the GeV excess seen in the Galactic Centre and Inner Galaxy to mitigate leakage of masked emission into the region of interest and point source confusion.

High Resolution Proton Radiography and Computed Tomography

Collaboration with Charles-Antoine Collins-Fekete

I assisted in developing a fast maximum-likelihood method to reconstruct projected images of relative stopping power from proton radiography data and in placing cubic spline charged particle path estimators into a Bayesian framework.

Undergrad Projects

Measuring the Lifetime of Trapped Sleptons using ATLAS

Supervised by James Pinfold, University of Alberta (Summer 2011)

In supergravity with the gravitino as the lightest supersymmetric particle, the next-to-lightest supersymmetric particle has a naturally long lifetime of hours to years. These long-lived particles could be produced at the LHC and then trapped in the detectors, with their decays being observed later (see Pinfold and Sibley, 2010). I started implementing a search for these decays in events not associated with collisions, using timing information to remove the downward-going cosmic ray background.

Nanomechanical Torsional Resonator Torque Magnetometry

Supervised by Mark Freeman and John Davis, University of Alberta (Summer 2010)

The Freeman group fabricated nanoscale torsional resonators to conduct magnetometry on nanoscale magnetic elements. I constructed a vacuum chamber apparatus to conduct optical interferometry measurements on these resonators at low temperature and adjustable magnetic field. This apparatus was used to observe the magnetic hysteresis of a magnetic film and the magnetic supercooling of vortex nucleation in a magnetic disk.

Evaluating Machine-Learned Protein Classifiers

Supervised by Paul Lu, University of Alberta (Summer 2007, Summer 2009)

The Proteome Analyst project was a publicly-available, web-based system for predicting protein properties using machine-learned classifiers constructed from labelled training data provided by the user. I created a web-based system to calculate classifier performance statistics by comparing a classifier’s output to known true classifications.