Before he could publish in Telugu, he had to coin the words

Siddhartha Jayanti, now an assistant professor of computer science at Dartmouth, has posted a paper on arXiv describing how he manufactured the Telugu technical vocabulary he needed before he could write up his own results ^[7]. The order of operations is the point: the words did not exist, so he derived them, and the derivation method is reusable across languages.

The underlying research is not linguistics. Jayanti analyzed how groups of processors coordinate by tracking what each one knows and how they communicate, and established fundamental mathematical limits on how quickly multiple cores can safely coordinate and share common data ^[2]. When he sat down to write it in Telugu, a language with roughly 100 million speakers ^[1], there was no settled term for something as basic to the work as "shared-memory multiprocessor" ^[3]. That is the barrier in concrete form: not a shortage of speakers, and not a shortage of results, but a missing lexicon standing between the two.

His fix was construction rather than transliteration. For "asynchronous" - which in distributed computing means processors work independently without waiting for other systems to stay in sync ^[13] - he built the Sanskrit asamakalika from a- (not), sama (same) and kala (time), then adapted it to asamakalikamu to fit Telugu grammar ^[12]. In the paper he notes that English got there the same way, assembling "asynchronous" from the Greek a-, syn- and khronos ^[14]. "Shared-memory multiprocessor" became samvibhakta-smrti bahusamsadhakamu, his most complex example, produced by modifying Sanskrit roots with prefixes and suffixes ^[15].

The reason to route through Sanskrit is leverage of a specific kind. Indian languages have long borrowed from classical sources with minor changes to word endings or sound substitutions, which Jayanti says means "you can create a common vocabulary across a slew of Indian languages all at once" ^[11]. That matters in a country of more than 1.47 billion people with dozens of languages, where Telugu is one of 22 official ones ^[8][9] and accounts for roughly 7 percent of the population by itself ^[20]. Jayanti also argues that natively derived terms teach: speakers can relate them to other words and understand them linguistically and scientifically ^[10]. A borrowed English string carries no such structure.

The second obstacle was mechanical. Keyboards were missing characters, fonts rendered Telugu incorrectly, and typesetting software could not handle the script ^[16], so he assembled a workflow he calls TeluguTeX to set Telugu text and mathematical notation correctly ^[17].

The work was not a side project that cost him standing. The Telugu chapter of his dissertation won the 2023 ACM Principles of Distributed Computing Doctoral Dissertation Award ^[4], the chapter was posted on arXiv in 2022 as the first modern computer science research paper in an Indian language ^[6], and its technical results have fed follow-on work in distributed and parallel computing, including efficient algorithms developed with collaborators at Princeton and MIT ^[19]. One detail is worth holding onto: the main results had to be translated into English for reviewers ^[5].

What to watch is whether the tooling and the terminology get adopted by anyone other than their author. Jayanti's own stated dependency is better fonts, keyboards and typesetting tools ^[18], which is a funding and standards problem, not a research one. The terminology test is whether Sanskrit-derived coinages spread to other Indian-language communities as he predicts, or stay a one-language artifact. And the reviewer question stands unanswered: a vocabulary is not much use if publication still requires an English version.