Weekend Projects in Applied ML
A few years ago I made a career pivot into quantitative finance. While I learned a wonderful amount of new terminology and economic concepts, I found myself continually returning to the techniques from my background in applied math, engineering, and statistics to solve this new class of problems. These techniques are so familiar to me because I spent years early in my career applying them to problems in painstaking detail, often by hand. Importantly, I consistently wrote up the results in a consice, technical manner for review. I cannot emphasize how often I found my own errors or unwarranted assumptions through the act of writing up the results.
Near the end of 2025, I resolved to do more technical writing to keep my skills sharp and continue learning in the way that has served me so well in the past. This simple website is the result of that resolution.
In our current age of AI-generated code and articles, I can’t help but wonder if this type of content will become less and less common. It takes a surprising amount of time to hand craft equations in LaTeX, double check derivations, generate figures, and write up a well reasoned discussion. But that repetition, reworking, and deep thinking is, in fact, what makes it such an irreplaceable tool for learning and understanding.
Highlighted Articles
Reinforcement Learning for Blackjack Strategy
This series was inspired by coincidence while reviewing some examples in Sutton and Barto’s Reinforcement Learning after just finishing Edward Thorp’s A Man For All Markets. I have long admired Edward Thorp after first becoming aware of his work while try to play advantage blackjack in the late 1990’s.
Tridiagonal Solvers
In graduate school I did work modeling heat transfer in biological tissue. In many cases the finite difference solution to the partial differential equations describing this model could be represented by triadiagonal matrix system. My solver at the time was purely sequential and single threaded. I wanted to see if I could revisit solution with modern parallel hardware and benchmark several algorithms for solving tridiagonal matrix systems.
