Tutored students in a variety of math courses, including single- and multivariable calculus, differential equations, and discrete mathematics, as well as the one-year physics sequence.

I wrote a dev blog post for the Azure Depth Platform team that describes the operation of Microsoft's indirect time-of-flight (ToF) system and some of the benefits of our implementation.

Lots of modern video games rely on the player's ability to hit the right inputs with sub-second timing, and the arrow-stomping DanceDanceRevolution is no exception. One of the long-standing problems that plagues its simulator community is accurately synchronizing hundreds or thousands of audio files to their step patterns - a task complicated by two competing historical "sync bias" paradigms: null sync vs. an offset of 9 milliseconds introduced by legacy software. Players often end up with a mix of null and +9ms patterns, which is a large enough difference to affect score and immersion.

I designed an algorithm that generates a beat onset response from the audio file, compares it to the map of beat onsets defined by the step pattern, and decides which bias paradigm it was meant to conform to based on the offset between the two. After confirming the results against popular "well-synced" patterns, I released a combo CLI/GUI program that offers visual representations of the sync quality and offset, batch processing, and a confidence metric. The audio signal processing is done in scipy, and the GUI makes use of matplotlib and wxPython.

The audio analysis is accurate enough (within 1-2ms for most files) that some early adopters have used the millisecond offset directly, not just to check the bias paradigm.

One of the long-standing problems that plagues the StepMania (DanceDanceRevolution simulator) community is measuring the difficulty of scoring well on a given step pattern. I proposed an algorithm that could estimate relative scoring complexity by comparing players' performances, and tuned it using the large dataset provided by the International Timing League's online tournament, held in the spring of 2022.

The algorithm primarily gauges the relative difficulty of each step pattern, but it also can rank players by overall skill level and identify scores that are strong or weak for that player. During International Timing League 2023, I expanded this algorithm to provide player-specific recommendations, identifying step patterns they had played where their score would be easiest to improve and providing a target score based on their overall ability.

A realtime implementation of Conway's Game of Life, rendered in-browser using twitch chat emotes as the live cell contents. Users can configure the canvas, resolution, and rate of the animation, then add it to their own streams. I originally implemented this as a VJ backdrop for the Club Fantastic Season 2 launch livestream.