Principal Speech Data Linguist

Innodata Inc.

9d•$160,000 - $185,000

About The Position

Innodata is a global data engineering company focused on enabling the responsible advancement of artificial intelligence. We provide data, evaluation frameworks, and human expertise to build trustworthy AI systems. Our services and platforms cater to Generative AI/AI builders and adopters, leveraging our 36+ year legacy of delivering high-quality data and outstanding customer outcomes. This principal-level role is for an applied expert counterpart to our Speech & Audio Research Scientist. You will own the linguistic standards and quality for high-volume segmentation and transcription workflows. Your responsibilities include defining linguistic standards, establishing quality frameworks, managing the quality lifecycle, designing human-in-the-loop workflows, handling complex linguistic cases, and collaborating with research scientists to translate model objectives into specifications. You will also train and mentor transcribers, and represent Innodata's approach to customers.

Requirements

Substantial industry experience (typically 8+ years) in transcription, segmentation, and speech-data quality, with a track record of authoring standards.
Bachelor's degree in linguistics, phonetics, or computational linguistics, or a closely related field.
Strong foundation in phonetics, phonology, and sociolinguistics.
Big-picture understanding of how transcription and segmentation choices impact downstream modeling.
Fluency in phonetic transcription and IPA.
Hands-on experience with acoustic and phonetic analysis of speech (spectrograms, formants, pitch, prosody, segment boundaries) applied to real-world data (e.g., in Praat).
Deep experience with audio segmentation conventions (utterance/turn boundaries, timestamping, speaker/diarization labeling).
Hands-on fluency with modern speech stack: Whisper, commercial ASR engines (AssemblyAI, Deepgram, Rev, Speechmatics), forced alignment (Montreal Forced Aligner), and annotation tools (ELAN).
Comfort scripting for speech-data work (Python for batch processing, QA, metrics like inter-annotator agreement and WER; regular expressions; Praat scripting).
Practical data-management skills across the delivery lifecycle (pre-processing, acceptance checks, post-processing, validation, reporting).
Multilingual capability and hands-on experience with accented, dialectal, and code-switched speech.
A point of view on the evolution of human-in-the-loop workflows as models improve.
Strong written and verbal communication skills.
Comfort working directly with research scientists and interfacing with customers.

Nice To Haves

Advanced degree preferred.
Low-resource languages experience a strong plus.
Responsible AI considerations for speech (bias, privacy, consent) a plus.

Responsibilities

Define transcription and segmentation standards, style guides, and annotation conventions (verbatim, clean/intelligent verbatim, IPA, phonetic transcription, timestamping, boundary segmentation, speaker labeling, diarization labels, disfluencies, non-speech events, code-switching, orthographic conventions).
Establish and run quality frameworks: rubrics, error taxonomies, adjudication processes, inter-annotator agreement, and human QA at scale.
Own the quality lifecycle for transcription and segmentation deliverables end to end: pre-processing, normalization, quality checks, post-processing, validation, and report creation/packaging.
Design human-in-the-loop workflows to optimize human review, correction, and adjudication based on ASR quality.
Handle linguistically challenging cases: accented/dialectal speech, low-resource/multilingual audio, overlapping speech, domain jargon, and noisy acoustic conditions.
Partner with Speech & Audio Research Scientist to translate model objectives into specifications and understand how transcription methods affect ASR, TTS, and speech/content-understanding models.
Train, calibrate, and mentor expert transcribers and reviewers, and build onboarding/calibration processes for scaling work.
Represent Innodata's transcription and segmentation approach to customers and contribute to methodology documentation.