Research Scientist, Speech & Audio
Core
Design, structure, and evaluate audio data and evaluation methodologies for speech and audio models (ASR, TTS, diarization, etc.) to ensure robustness across accents, noise, and code-switching.
Role type
Senior Research Scientist (Speech & Audio)
Builds
High-quality audio datasets, evaluation benchmarks, and experimental validation frameworks for frontier labs and customers.
Domain
Artificial Intelligence, Speech & Audio, Data Engineering
Deliverable
production ML models
Required skills
Speech/Audio ML (ASR, TTS, diarization), PyTorch, dataset design, experimental validation, adversarial evaluation, multilingual/accented speech expertise, publication record
Preferred skills
PhD in relevant field, responsible AI/red-teaming experience, open-source contributions
Technologies
ESPnet, NeMo, SpeechBrain, Kaldi, HuggingFace, forced alignment
Responsibilities
Define data specifications and evaluation criteria for speech models; build evaluation methodology beyond WER (semantic accuracy, DER, naturalness); structure and sample audio data across diverse conditions; partner with engineering to scope audio capture programs; run experiments to validate data decisions; design adversarial evaluations; publish benchmarks and papers.
Seniority
Senior, hands-on IC
