Above is a diagram of all the musical recordings the Library of Congress deemed aesthetically, culturally, or historically significant: when they were recorded, what genre they fit in (or this music fan's best choice), and how long they play. Included are hundreds of songs, albums, and entire collections, such as the Bristol recording sessions (the yellow spike in 1927) or the Crescent City Living Legends collection (red spike in 1973). These data, as well as when the Library of Congress deemed the recordings worthy of inclusion, how the recordings can be accessed, and where all of this information comes from, form a PowerBI (a live-updating dashboard) the static form of which proves how beautiful data can be.
Besides PowerBI, a visualization tool which depends heavily on Excel, I've done data analysis with LabVIEW and extensively with MATLAB, and though this is largely for routine and less impressive coursework, two exceptions are described below.
In the summer of 2021, a then-remote student and I attempted to evaluate the previous unmeasured thermal conductivity of a rare substance that could potentially be used for heat energy storage in high-temperature aerospace applications. Measuring the property - a function of temperature - at the desired 1000 C of the air-reactive substance proved difficult and was not managed by the end of the summer, but a fair amount about formula fitting, experimental set-up, and data recording was learned.
ENGR 132, a core curriculum course for all Purdue engineering students, was dominated by a large data analysis project. The project began with a .csv file of over 500,000 points of data representing the concentration of various substrates in time as they were dissipated by various enzymes. These data formed curves that somewhat resembled those delineated by the Michaelis-Menten model of enzyme kinetics, but were filled with noise suggesting randomly inaccurate measurement equipment. Below is the final result of the team project, a technical brief outlining how we met these challenges; if you're unusually curious about the original data, it is available here as a 1,920 page pdf.