Research Scientist, Speech & Audio

This is a fully remote position, open to applicants in United States.

📋 Description

• Develop the design, structure, and assessment frameworks for audio data utilized in speech and audio models.

• Convert requirements for speech and audio models into detailed data specifications, including modalities, transcription and annotation schemas, sampling methods, and evaluation standards.

• Create evaluation methodologies that address semantic accuracy, robustness to noise and accents, code-switching, diarization error rates, speech naturalness and intelligibility, as well as streaming latency.

• Organize, enhance, and sample audio from diverse languages, accents, acoustic environments, demographic groups, emotional tones, paralinguistic ranges, scripted versus spontaneous speech, and various speaker configurations.

• Collaborate with transcription and linguistics leaders to formulate transcription specifications and assess their influence on model performance.

• Work alongside audio solutions and engineering teams to guarantee that collected audio aligns with model objectives.

• Conduct fine-tuning and evaluation experiments with ablation studies that correlate data selection with quantifiable improvements.

• Design adversarial tests and challenging evaluations to discover failures in speech systems.

• Publish benchmarks, methodologies, and research findings.

• Collaborate with annotation teams, subject-matter experts, and synthetic- and augmented-audio pipelines to implement specifications effectively.

• Engage directly with customers and leading labs that are developing ASR, text-to-speech, speech-to-speech, conversational voice, diarization, and audio-language models.

⛳️ Requirements

• Approximately 5+ years of practical experience in the speech or audio machine learning industry.

• A Bachelor's degree in computer science, electrical engineering, or a related technical or quantitative discipline is mandatory.

• Experience in training and evaluating speech or audio models — including ASR, TTS, speech-to-speech, speaker, or audio-language models — with a solid foundation in PyTorch.

• Proficiency in ESPnet, NeMo, SpeechBrain, Kaldi, HuggingFace, forced alignment, WER/CER, and other relevant speech metrics.

• Practical experience with multilingual, accented, dialectal, low-resource, or code-switched speech.

• Familiarity with synthetic or augmented audio, encompassing TTS pipelines, noise simulation, and room-response modeling.

• Experience in creating evaluation sets and analyzing dataset coverage across various conditions.

• First-author publications or significant contributions to open-source projects in venues such as Interspeech, ICASSP, ASRU, SLT, or NeurIPS.

• Capability to collaborate directly with research scientists from customer and frontier labs.

• Ability to clearly communicate data and modeling choices to both expert and lay audiences.

• A methodical and reproducible approach to experimentation and documentation.

• Interest or practical experience in responsible AI evaluation and red-teaming is an advantage.

🏝️ Benefits

• Competitive salary and performance-based incentives.

• Comprehensive health, dental, and vision insurance.

• Opportunities for professional development and career advancement.

• Flexible work hours and remote work options.

• Collaborative and innovative work environment.