Senior Speech & Audio Data Scientist - Benchmarking

Innodata Inc.; United States; On-site; USD 160,000 - 185,000; Full time

Job summary

Innodata is hiring a Research Scientist to own the data and evaluation science behind speech and audio models used in ASR, TTS, and conversational voice.

You will define data specifications, transcription schemas, and sampling strategies, then build robust evaluation methods that go beyond word error rate to measure robustness, latency, and naturalness. You’ll collaborate with customers and labs to shape data to objectiv es.

Qualifications

• 5+ years of hands-on industry experience in speech or audio ML.
• Bachelor's degree in CS, EE, or related field; MS/PhD preferred.
• Experience training and evaluating ASR, TTS, or audio-language models with PyTorch.
• Fluency with ESPnet, NeMo, SpeechBrain, Kaldi, HuggingFace and related metrics.

Responsibilities

• Define how Innodata designs, structures, and evaluates audio data for speech and audio models, and validate those choices experimentally.
• Translate model requirements into concrete data specifications: modalities, transcription/annotation schemas, sampling, and evaluation criteria.
• Build evaluation methodology beyond WER to include semantic accuracy, noise/accents, DER, naturalness, and streaming latency.

Skills

Speech ML

PyTorch

Audio datasets

Benchmarking

Education

Bachelor's degree in CS/EE or related field

Tools

ESPnet

NeMo

SpeechBrain

Kaldi

HuggingFace

WER/CER metrics

Job description

Innodata is hiring a Research Scientist to own the data and evaluation science behind speech and audio models used in ASR, TTS, and conversational voice.

You will define data specifications, transcription schemas, and sampling strategies, then build robust evaluation methods that go beyond word error rate to measure robustness, latency, and naturalness. You’ll collaborate with customers and labs to shape data to objectiv es.