Computer science, at every scale.

About & data

How the atlas is built

Every observation points back to a source. None of the computer science inventories is yet certified complete.

Method in brief

• Universities. The 2025 Academic Ranking of World Universities, through rank 300 (ranking ↗). At Princeton, CMU, EPFL, MIT, Berkeley, Stanford and Michigan the atlas shows each computing department’s PI labs, one per faculty PI, when the lab has at least one lab paper. Elsewhere it shows computer science units with at least one linked paper. Units without papers remain in the collection and are not shown; each university lists how many are hidden. Michigan data is incomplete: student records cover 117 of 287 labs, 89 link checks remain open, and stored copies cover about 56% of papers.
• Research units. Labs, groups, centers, institutes and chairs come from official institutional sources. An undated directory listing establishes a candidate unit, not its active years.
• Publication views. PI as last author: a verified author identity is last on the work and that authorship lists the university; this is a PI-based proxy. Reviewed lab bibliography: an explicit lab bibliography supports the link, regardless of author order; this confirms a bibliography listing, not historical lab tenure. The combined view is their union. None is exhaustive.
• History. Dated evidence supports its stated year or event interval. Years between observations stay unknown; no lifespans are invented. Counts are observed lower bounds.
• Attribution. A lab paper needs a verified PI author on that exact paper, evidence of the PI’s faculty appointment at publication, and a supported connection to the lab. Coauthorship, rosters and advising are kept as separate relationships.

Data

• PI labs of the CS departmentsCSV
• The 300-university cohortCSV
• Computing coverage, 300 universitiesCSV