Arabic Linguist (Speech and Localisation)

Own how our AI speaks and writes Arabic. You will set the standard for Emirati dialect accuracy across our speech systems, for Modern Standard Arabic across our product interfaces, and for the quality of every audio and transcript corpus we train on.About Xpin AIXpin AI builds sovereign AI for governments, enterprises, and institutions across the GCC. Our products are bilingual at their core, and Arabic is not a translation layer applied at the end, it is a first-class product requirement. Our speech systems are expected to sound natural to Emirati users, and our interfaces are expected to read correctly to government readers. The roleYou are accountable for Arabic quality, you will work between the AI engineering team building our speech models, the product designers building our interfaces, and the data pipeline that feeds our training.The role spans two registers deliberately. Our interfaces need Arabic at a formal, government/enterprise appropriate standard. Our speech systems need Emirati Arabic that sounds like a native speaker rather than a generalised Gulf accent. You need sound judgment in both and the discipline to keep them separate.You will establish a terminology base and Arabic style guide the whole company works from, a documented evaluation method for Emirati speech output that gives engineers a repeatable signal, and a corpus quality process that catches problems before they reach training rather than after.What you'll ownEmirati dialect accuracy in speech systems. Evaluate TTS output for pronunciation, prosody, stress, intonation, and naturalness. Identify precisely why a sample sounds wrong, in terms an engineer can act on, rather than reporting that it sounds off.Text normalisation and pronunciation rules. Define how numbers, dates, currency, times, abbreviations, acronyms, proper nouns, and English loanwords should be spoken in natural Emirati Arabic. This is one of the highest-impact and most overlooked parts of building a TTS system.Training corpus quality, audio and text. Vet recordings and transcripts for dialect authenticity, transcription accuracy, speaker consistency, audio quality, and coverage gaps. Define transcription conventions and hold contributors to them.Annotation standards and reviewer management. Write the annotation guidelines, train and calibrate reviewers and voice talent, measure agreement between annotators, and resolve disputes with a documented rationale.Product interface Arabic. Own translation and UX writing quality across our products, in the correct register, working with the design team on how Arabic behaves in a right-to-left interface.Terminology and style. Build and maintain the Arabic glossary and style guide covering technical, AI, and government terminology, so the same concept is not rendered three different ways across three products.Linguistic support to engineering. Advise on phoneme inventories, diacritization, grapheme-to-phoneme handling, code-switching between Arabic and English, and evaluation design for Arabic speech and language models.What we need you to knowEmirati Arabic at native or near-native level. You do not need to be Emirati. You do need to speak, hear, and judge Emirati Arabic as a native listener would, and to distinguish it reliably from Kuwaiti, Qatari, Bahraini, Saudi Najdi, and Omani speech. Generalised Khaleeji competence is not sufficient for this role.Modern Standard Arabic at professional writing standard, at the level expected in formal government communication.Arabic diglossia in practice. Confident judgment about which register belongs where, and awareness of variation within the UAE itself, including differences between Abu Dhabi, Dubai, and the Northern Emirates.Arabic script and typography in software. Letter shaping and connection, bidirectional text, Eastern Arabic-Indic and Western numeral conventions, Hijri and Gregorian dates, and Arabic plural rules, which have six categories and routinely break interfaces built for English.Analytical rigour. You can turn a subjective impression into a written, testable, repeatable criterion. This is the difference between reviewing language and owning it.Tools and technical proficiencyLocalisation: a CAT and translation management platform such as Crowdin, Lokalise, Phrase, memoQ, or Trados; translation memory and terminology base management; ICU message format and pluralisation.Phonetics and audio: Praat for acoustic and prosodic analysis, ELAN or similar for time-aligned annotation, comfort listening critically at scale.Transcription and annotation: annotation platforms such as Label Studio or Prodigy, orthographic and phonetic transcription conventions, inter-annotator agreement measurement.Arabic linguistic tooling: diacritization tools, Buckwalter or IPA transliteration, grapheme-to-phoneme concepts, morphological analysis.Evaluation: designing and running listening tests such as MOS and preference comparisons, and reading engineering metrics such as word error rate well enough to interpret them.General: spreadsheets at a high level, regular expressions, and basic Python or scripting comfort is a strong plus for working with corpus data at scale.Nice-to-haveExperience on a TTS, ASR, or voice assistant product, particularly for Arabic or another diglossic language.Formal background in linguistics, phonetics, translation, or computational linguistics.Experience recruiting, directing, or managing voice talent and recording sessions.Managing distributed annotator or transcriber teams.Prior work with UAE government entities and familiarity with their language conventions.English at professional working level, which you will need daily with an international engineering team.Other regional dialects, particularly wider Gulf varieties.Why joinArabic AI is genuinely underserved, and Emirati Arabic more so. Most speech systems available today either speak Modern Standard Arabic in a way no one actually talks, or apply a generic Gulf accent that Emirati users hear as foreign. You will be defining how this is done properly, with direct influence on models that ship into real government and enterprise environments.This is a rare role in that your judgment sets the standard rather than reviews someone else's. You will work alongside AI engineers who need your expertise and will act on it, in a company where Arabic is treated as core product rather than as a compliance requirement. Compensation and logisticsOn-site in Dubai. Compensation is competitive and benchmarked to your experience and the UAE market. The package includes UAE employment visa sponsorship, medical insurance, and annual leave and flight allowance per UAE labour law and company policy. Xpin AI hires on demonstrated ability and welcomes applicants of all nationalities.How to applySend your CV along with a short note, in Arabic, on a language quality problem you have solved. If you have worked on speech or voice products, tell us what was wrong with the output and how you diagnosed it. Shortlisted candidates will complete a practical assessment involving Emirati dialect evaluation and corpus review.