Senior Data Scientist, Audio

Role overview

A senior data science role focused on conversational audio, sitting at the intersection of research, engineering, and data operations. The work centers on understanding what makes real-world speech data difficult — accents, dialects, overlapping talk, code-switching, domain vocabulary, and varied acoustic conditions — and turning that understanding into measurable model gains. This is a hands-on, high-leverage position with latitude to define an area from first principles.

Responsibilities

- Characterize large-scale conversational audio corpora across languages, acoustic conditions, domains, speaker demographics, and quality, and identify gaps where representation is thin. - Design and operationalize active-learning loops that systematically decide which data to prioritize next based on expected model impact. - Architect human-in-the-loop workflows, tooling, and model-assisted steps that maximize the value of limited human attention. - Build curated benchmarks and methodologies that allow honest, representative claims about model quality across diverse real-world speech. - Convert scrappy proofs of concept into repeatable, documented, production-grade pipelines that other teams can run independently. - Partner closely with research and engineering to translate findings into model and product improvements.

Requirements

- Hands-on experience on real data pipelines and model-facing problems in data science, machine learning, or applied research. - Strong Python and general data tooling skills, with comfort building analysis, scoring, and automation directly. - Practical experience with data characterization, data selection, active learning, or similar prioritization problems. - Working familiarity with speech/audio or NLP models, including reasoning about output quality, confidence, and error modes. - A track record of turning ambiguous, messy data situations into measurable improvements. - A bias toward building reusable systems rather than one-off notebooks, plus strong communication skills for translating complex findings to non-specialist audiences. - Active, hands-on use of modern AI tools integrated into daily workflow.

Nice to have

- Direct experience with automatic speech recognition, text-to-speech, multilingual or code-switched audio data. - Experience with ensemble labeling, pseudo-labeling, or LLM-assisted annotation. - Familiarity with data provenance, PII and GDPR-aware pipelines, or model-improvement compliance. - Experience building custom or fine-tuned models for specific domains. - Comfort working alongside research and engineering on shared infrastructure.