I got bit by the programming bug in my teens, when my dad brought home a hand-me-down from work: a 486DX2 running Windows 3.1 at 50MHz with 4MB of RAM. It had some (agonizingly slow) games and productivity software on it, but I wasn't truly hooked until I pirated a copy of Turbo Pascal from the school computer lab and found out you could make *anything*. After I'd been studying computer science for 2 years at university, one of my professors gave me my favorite piece of professional advice: study something else too, because a hard-sciences engineer or physicist, who is also an expert in programming/automation, is far more purposeful and effective than either skillset taken alone. So I starting picking up EE electives, got hooked on Maxwell's equations, and have been programming computational electromagnetics software ever since.Today I author design automation tools for all sorts of EM phenomena: antenna radiation, radar scattering, optical diffraction, etc. Typically these are frequency-domain modelers based on finite elements (FE), hybridized with some sort of boundary integral equation (BI). But I appreciate methods of all shapes and sizes, from AIM to Z-Matrix, and think they've all got their place/problem where they excel.The most influential code I authored is probably SENTRi, a suite of antenna-on-platform modelers based on FE+BI and FE+DDM, that the DoD offers as government furnished equipment to run on their supercomputers. You might have even heard of this one, because a lot of defense companies list SENTRi experience as a desirable qualification for signature management engineers. MIRAGE is another one, an FE+mode matching solver for optical/IR scattering from periodic structures (often metalenses, possibly nonlinear). But my best product.. has yet to come. Or at least that's how I like to think about it. Each one ought to do something a little different/better than the last.In my free time I run a small business (MyraCore) whose principal product is an open source sparse-direct solver (MyraMath) that has specialized Schur-complement/partial-solution functionality to support all of these FE+X modeler concepts. Of course, high performance sparse solvers depend crucially upon high performance dense solvers, so I've also done my share of BLAS/LAPACK tinkering, which is encapsulated in a kernel-library (MyraKL) that's also open source.