Linguist – Tamil, Marathi or Egyptian Arabic languages-SUC-SMN-001

Contract Duration: 6 Months (Temp to Hire)

Summary:

• The main function of a TTS Linguist Contractor is to determine speech data needs and make for data-based model and product improvements.

Job Responsibilities:

• Provide linguistic expertise in the areas of phonetics, phonology, lexicography, dialectology, and NLP.
• Design and conduct experiments for evaluating transcription quality.
• Develop manual and automated processes for multiple concurrent projects including ensuring high-quality label alignments, prosodic classification, POS identification and disambiguation, targeted modeling data, and user feedback.
• Create and perfect text normalization and inverse text normalization processes.
• Manage lexical and phrasal transcriptions and related metadata.
• Analyze system metrics such as user opinion, lexicon transcription coverage, and POS tagger performance and remedy pain points.

Skills:

• Knowledge of phonetics, phonology, sociolinguistics, dialectology, and other areas of linguistics.
• Ability to analyze waveforms and spectrograms.
• Knowledge of prescriptive writing and punctuation conventions for at least one language.
• Excellent communication skills both verbal and written.
• Knowledge in transcription and annotation systems such as SAMPA, IPA, and ToBI.

Education/Experience:

• Bachelor’s degree in linguistics, language technologies, computational linguistics, speech science, or related field.

Top 3 must-have HARD skills:

• Native-level fluency in Native-level fluency in Tamil, Marathi or Egyptian Arabic (+ familiarity with Modern Standard Arabic is nice to have).
• Familiarity with command-line, scripting, and versioning systems.
• Phonetics/phonology, including experience with transcription in IPA or SAMPA -OR- experience with regexes.

Good to have skills:

• Python.
• Data analysis.

Core Responsibilities of TTS:

• Build and improve TTS models (speech generation) to sound more natural, expressive, and robust, including things like prosody and non-verbal cues (e.g., laughter/breath).
• Multilingual + i18n support: expand language coverage and handle tricky cases like code-switching, accents, and language ID.
• Deploy/integrate models into products (sometimes including on-device inference constraints for wearables).
• Evaluation + quality measurement: develop pipelines/guidelines to measure naturalness and expressivity.
• Native speaker expertise for new TTS locales.