Physics and astrophysics aren’t the only things I’m good at. My time at UC Irvine and before has taught me a wide variety of skills that can apply to many areas.

Coding

Python is my main coding language, and where I feel most comfortable. Quite a lot of astrophysics packages are optimized for python, and so it is a natural choice. Common python tasks that I perform daily include making publication-worthy plots, fitting models to data, and cleaning data using a variety of methods.

I am familiar with R, as I have had to use it to run, mainly, statistical software packages.

Similarly, I am proficient in C++, as I have had to run models in C++, and have had to troubleshoot a variety of issues.

Data Analysis

Observational astronomy and astrophysics requires extensive data analysis. Data can come in a wide variety of forms. The data I analyze ranges from 2D image data, spectroscopic wavelength data, 1D timeseries, and much more! These data can be taken by ground based observatories, spacecraft, or telescopes in someone’s backyard! During my internship at Lawerence Livermore National Laboratory’s (LLNL) Data Science Summer Institute, I learned to deal with even more kinds of data: laser physics data from experiments, and chemical data predicting the binding affinity of ligands.

Consequently, I have a lot of experience cleaning data, and especially identifying what problems may exist, and how to fix them. In astronomy, data contamination happens all the time. A spacecraft orbiting Earth may experience glare from the Sun while studying a star. Temperature fluctuations in a spectrometer here on Earth might introduce a systematic offset that affects all data it takes. A tired observer may have notated the wrong target at 4 am. Errors happen all the time, and fixing them is no less important than modeling the data.

I use a wide variety of modeling techniques. Different datasets require different analysis methods. A common technique I use when modeling exoplanets is Markov-Chain Monte Carlo (MCMC) sampling to measure model parameters when studying exoplanets. I’ve tied this into model comparisons and periodogram analysis. I often combine this with Gaussian Process regression, or other sophisticated likelihood estimators. My time at LLNL forced me to learn even more machine learning methods such as neural networks and random forests for classification, and even combining several methods into an ensemble to improve classification performance.

Statistics

It is impossible to do proper physics without a strong understanding of statistics. My background in frequentist methods is strong, but my real expertise is in the Bayesian approach to modeling and data. I have handwritten and debugged likelihood functions, prior functions, and sampling routines. Assessing model probability and comparisons is a frequent task that I perform here at UCI.

High Performance Computing

I am a regular user of UC Irvine’s HPC cluster. Using a slurm scheduler, I frequently run thousands of hours of jobs, batched or parallelized, in order to analyze datasets that would normally take weeks–or even longer.