Research Scientist, Speech & Audio

Overview

Innodata is hiring a Research Scientist to define the science behind speech and audio data design, evaluation, and model improvement. This remote role is focused on building and validating the data and evaluation frameworks used for ASR, text-to-speech, speech-to-speech, conversational voice, speaker diarization and verification, audio-language models, and streaming systems.

You will work directly with customers and frontier labs, partnering with transcription, linguistics, audio solutions, engineering, annotation teams, and subject-matter experts to turn model objectives into concrete data specifications and operational plans.

What You'll Do

• Translate speech and audio model requirements into data specifications, including modalities, transcription and annotation schemas, sampling plans, and evaluation criteria.
• Design evaluation methods that go beyond word error rate, including semantic accuracy, robustness to noise and accent, code-switching, diarization error rate, naturalness and intelligibility of generated speech, and streaming latency.
• Determine how audio should be structured, enriched, and sampled to achieve coverage across languages, accents, acoustic environments, speaker demographics, emotional and paralinguistic variation, scripted and spontaneous speech, and single- and multi-speaker settings.
• Work with transcription and linguistics leads to convert model goals into transcription conventions and measure how transcription quality affects ASR and speech model results.
• Partner with audio solutions and engineering teams to specify what good data and evaluation require for collection programs.
• Run experiments that test the impact of data decisions by fine-tuning and evaluating models on Innodata data, including ablation studies tied to measurable improvements.
• Design adversarial and stumping evaluations using noisy, accented, and adversarial audio to identify failure modes and improve data coverage.
• Work with annotation teams and the synthetic- and augmented-audio pipeline to turn specifications into execution plans.
• Publish benchmarks, methodology, and research papers that contribute to the field and support customer trust.

Requirements

• About 5+ years of hands-on industry experience in speech or audio machine learning.
• Bachelor's degree in computer science, electrical engineering, or a related technical or quantitative field is required.
• Advanced degree such as an MS or PhD in a relevant field is preferred.
• Experience training and evaluating speech or audio models such as ASR, TTS, speech-to-speech, speaker, or audio-language models.
• Strong PyTorch fundamentals.
• Familiarity with speech toolchains and metrics such as ESPnet, NeMo, SpeechBrain, Kaldi, HuggingFace, forced alignment, WER/CER, and related evaluation measures.
• Hands-on experience with multilingual, accented, dialectal, low-resource, or code-switched speech, plus synthetic or augmented audio such as TTS pipelines and noise or room-response simulation.
• Experience building evaluation sets and reasoning about coverage across conditions for speech data.
• First-author publications or strong open-source contributions at venues such as Interspeech, ICASSP, ASRU, SLT, or NeurIPS.
• Ability to communicate clearly with research scientists and non-expert stakeholders using rigorous, reproducible experiments and documentation.

Preferred Experience

• Interest or hands-on experience in responsible AI evaluation and red-teaming, including spoofing, voice-cloning robustness, or bias across accents and languages.

Compensation

The expected salary range for this role is $160,000 - $185,000 per year, depending on experience, skills, and qualifications.