My field: computational astrophysics, with a multi-messenger background. The question that has focused my career is how experimental design and the process of data analysis shapes the theories chosen and how the theories developed shape the results of data analyses.
I began to frame this question in college through discussions with a blind friend, who challenged my perspective that the world had an underlying reality with his firm belief that the world was constructed through our perceptions of it. The question I posed to him, was how computers originated from the underlying physical reality. In my undergrad thesis in the Center for Space Research, I used numerical computational methods based on mathematical algorithms approximated for the purpose of use on a digital computer to predict what a specific data analysis and measurement technique would reveal, for a hypothetical larger sample of gravitational lens systems with one dimensional velocity dispersions and Einstein ring radii drawn from a parent distribution based on a handful of measurements of known similar systems. Moving on to my first masters degree at Ohio State, I considered another question of measurability and data analysis using computational physics, where I modelled the observable population of planet transits around the observable population of stars in the OGLE-III lensing survey, and based on the magnitude limits of the survey and the real population of detected planets, determined the proportion of stars with hot and very hot Jupiters along two different lines of site in the modelled (and real) galaxy. To be fair, there was quite a bit of work done following my time on the planet paper. I moved on to University of Minnesota to pursue a curiosity about what at the time was called particle astrophysics (rather than multi-messenger astronomy) and also to be closer to family.
At University of Minnesota, the projects available to me were not quite what I recalled from my experience over the summer as an undergrad. As an undergrad, there were cosmic ray data analyses being done in the MINOS neutrino detector. However, those were no longer current or relevant by the time I was a grad student, and instead the new focus was on the NOvA detector being built off the axis of the neutrino beam from Fermilab. As an undergrad, I did a very short project modelling the momentum versus energy of the tracks in the detector using a MINOS specific C++ framework. When I returned as a grad student, though, what they really needed was someone who could monitor some Avalanching Photodiodes as they cycled through cooling and reheating, to make sure they were robust enough for the northern Minnesota winter to use in NOvA. I cannot say experimental particle physics was really my thing, and I was grateful to lose my funding and move on, although I loved the classes I took and passed my general exam in that field. I had a summer research fellowship in theory, where I computed a component of the anomalous muon magnetic moment using form factors to tree level, successfully, as asked.
I probably would have stayed in the field if my communication with my advisor had been better, but I had hearing loss, and he muttered. He wasn't the first advisor who had that issue-- actually, I couldn't understand Walter Lewin the summer I worked for him at MIT either! It was basically a miracle that I had any idea what my project was in either of the two summers. I nearly failed junior lab for that reason as well, and had to take an oral exam on paper. Sorry people, if I don't know what the question is, I really can't do what you're asking, you have to speak up.
I was thrilled when, that fall, a new advisor came to University of Minnesota in gravitational waves. The term "multi-messenger astronomy" really had not been invented yet, but I could see the obvious potential. So could other people, of course. He also had data analysis and experiment-modelling projects available to me, and it sounded like the actual amount of physical lab work I would need to do would be minimal or none. PERFECT. The very first fields that had interested me in high school was gravity probe B (I had done a project about that in a summer camp), and the second thing that had inspired me was the major supernova breakthrough for dark matter and dark energy that won the Nobel Prize (I had seen a talk about that at Physics Olympiad), and I had also been majorly inspired by a course at MIT called Exploring Blackholes, another on the Inflationary Universe, a project I had done on the Solar Neutrino Problem at MIT, and some particle astro classes at Ohio State. I had loved my General Relativity course at MIT and my Cosmology course at Ohio State. I was enrolled in GR and Cosmology at UMN. It was a great fit. And, LIGO. I had wanted to work on it at Caltech, twice, and been turned down, twice (although I did have the SURF to work on something else the first time and the REU to work on LIGO at Hanford the second time, but chose something else both times because it wasn't quite the right fit). So, I was extremely grateful for the third chance, although it was at UMN (which I still loved at the time-- I had chosen UMN because it was where I wanted to be, other than, I suppose, Caltech or something). (Okay so what I really wanted all along was to do data analysis for LIGO at Caltech... being totally honest... but I think it is more complicated than that...) (By the way thank you to those who found me a place along my career path and made this plan work out!!!)
Okay so returning to that... The neutrino project was basically given to someone else. Fine. He was a friend. I saw him again at LIGO Livingston.
Blah blah stupid shit
But lots of statistics to make the search better and the noise cancellation better
Some modelling of signals
Some detailed tracing of code, why exactly does this work how?
Same sort of stuff I did at LSU again actually, although from a theory point of view instead of experiment, which was really perfect in retrospect.
At LSU I also had the opportunity to follow along on discussions of LIGO detector characterization-- which is not really a skill I have-- though I would call my field similar in the sense that detector characterization links instrumentation to data analysis, and my skills/techniques have typically linked data analysis to the modelling of theory in computation and to experiment design. So perhaps my strengths could be called analysis characterization or something like that, rather than detector characterization, though I have viewed it from all sides.
At LSU I also had the opportunity to work a little bit with computational techniques such as learning further numerical mathematical methods for computing both in research and in classes, as well as learning a little bit about supercomputing, databases, and a small amount of other computer science concepts. I made extensive use of proporties of roundoff and truncation noise, evaluating convergence with variations in sizes of the system in various features such as time step, discontinuous galerkin nodes, or multipole moments. I also learned a little bit about the properties of what leads to more or less speedup under different situations, although I am not sure I could model it in detail or implement it right now. (I also don't have a supercomputer allocation or mega-dollars for something like AWS). Beyond LSU, I have studied machine learning and deep learning, and I can very clearly see the mathematical connections to the physics and statistics inherent in the computational methods themselves.
Beyond LSU I have also continued with orbits through a 3 body code and following rocketry forums, and also followed some quantum gravity forums (which I very preliminarily began to learn at LSU). I can see that quantum gravity would be a good direction to take a hobby interest in the future, in addition to multi-messenger astronomy, and I have some questions I would like to answer for myself (or have someone else think about) there.
However what I would really like is to turn this into a paying career. If you need a multi-messenger astronomy computational physicist specializing in "analysis characterization" in the sense that I understand the interface between data analysis, experiment design and background noise modelling, and theoretical prediction in a computational setting, please let me know.