rs.ui is a package that I created that edits the RStudio IDE theme to be completely customized which is not something that can typically be achieved. Here is an example of a theme generated by my package: It can either generate a theme given one color value (if you have a favorite color, for example) or given a csv file specifying the exact colors to change. This way, users can easily share and export their favorite themes instead of modifying css files. Here's the GitHub for the package.
Spatial Clustering:
This project uses OSRM (open source routing machine) to obtain drive times to and from census blocks and cluster off of a distance matrix of these drive times instead of Euclidean distances. A custom algorithm was built to equalize projected utilization within each cluster as existing clustering algorithms were failing in that regard. Discrete event simulation (DES) was used to add realism, simulating weekends, weather, etc. An R Shiny application was built using leaflet and highcharts as the main drivers; I pushed this application to my open source shiny server (configured with Digital Oceans). I also wrote corresponding blog posts for each step of the project.
I was able to import and use my Apple Health data in R, comparing my heart rates during and outside of mindfulness session, mapping a hike I recently took in Curt Gowdy State Park, and demo-ing how to trend activity summaries such as standing hours or steps.
My graduate project advisor recently reached out to me with the opportunity to get involved with a shiny app that plugs into Eliteform and VALD workout monitoring systems. I used the shinyMobile package to specifically design the app for usage on phones. To backup/pull data, I've employed cron-scheduled scripts on my Ubuntu 20.04 server hosting my shiny apps. Every 15 minutes, data from the API's are pulled, cleaned, and placed into the real shiny app's directory for use. This was a fun project as I learned how to use a server instance to deploy shiny apps, secure traffic with HTTPS using nginx, schedule data pulls for apps, and even add authentication to the real app.
My graduate project involved predicting Covid sample lineage identification success using coverage rates of base-pair regions. I used Random Forest to deem which base-pair regions were important to successful lineage identification and was able to predict whether a sample could be correctly identified or not with ~96% accuracy. My advisor on this project recently gave a presentation about it in Europe and we hope to have a journal article submitted within the year. Though work has continued on this project, my original paper and video of myself presenting are below:
Most of my other small projects/interests can be found on my blog. I'm planning on a post demonstrating how to import Apple Watch data into R as well as a post on plugging into the Spotify API within R.
