Research with Cal Poly's Kinesiology & Public Health Department and a World Health Organization-aligned consortium (Cal Poly, NCI, Oxford, Monash), testing how accurately wrist-worn device algorithms measure physical activity. Aug 2026 – Present.
Wrist-worn devices like smartwatches and research accelerometers are used everywhere to track activity, but different algorithms can give very different answers for the same person. The WAVES consortium pools validation data from research groups around the world and runs a set of open-source wrist algorithms against it, to see how well each one estimates moderate-to-vigorous physical activity (MVPA), steps, and sedentary time. The end goal is open, verified standards so researchers and clinicians can trust the activity numbers coming from these devices.
The work brings together public health, kinesiology, and data science teams, including Cal Poly, the National Cancer Institute (where activity and sitting time matter for cancer risk and survivorship research), the University of Oxford, and Monash University.
One of the validation datasets comes from a UWM study where each person's steps were coded by direct observation. I wrote the Python code that turns those raw observation files into second-by-second data the WAVES pipeline can compare against the algorithms.
The main WAVES repository is an R package built with targets, renv, Quarto, and pkgdown that runs every algorithm on a site's raw accelerometer data and summarizes the results. I'm one of six contributors and have merged algorithm fixes into it. The methods it currently runs: