Senior Data Scientist, Data Flywheel

Deepgram•Remote, CA

2d•$165,000 - $220,000

About The Position

Deepgram is seeking a Data Scientist to bridge the gap between research and data, focusing on the intricacies of conversational data. This role involves understanding the composition, value, and optimal utilization of conversational data. The position presents significant scientific, engineering, and infrastructure challenges due to the complexity of speech data, which includes diverse speakers, accents, dialects, emotions, overlapping speech, code-switching, domain-specific vocabulary, and varied acoustic conditions. The successful candidate will collaborate with research, engineering, and data teams to rigorously analyze and scale solutions for these challenges. This role requires a deep understanding of what makes conversational data difficult, the ability to characterize it, and the skill to shape strategies that translate data insights into model improvements. The position rewards conviction, creative approaches, and a willingness to challenge assumptions based on data. It is a hands-on, high-leverage role with considerable autonomy to define an area from foundational principles, reporting to the VP of Data Operations. Ideal candidates question established methods, can design efficient experiments, creatively identify signal sources (metadata, model confidence, failures), scale initial concepts, and leverage AI to enhance their own productivity.

Requirements

Hands-on experience with real data pipelines and model-facing problems in data science, ML, or applied research.
Strong proficiency in Python and data tooling, with the ability to build analyses, scoring, and automation independently.
Experience with data characterization, data selection, active learning, or similar 'what should we work on next' problems.
Working familiarity with speech/audio or NLP models, including the ability to reason about model output quality, confidence, and error modes.
A proven track record of transforming ambiguous, messy data situations into measurable model or product improvements.
Ability to build reusable systems that function autonomously, rather than one-off notebooks.
Strong communication skills, particularly in translating complex findings for non-technical audiences.
Active and demonstrated use of AI tools, including experience building with them.

Nice To Haves

Direct experience with ASR/TTS, audio data, or multilingual/code-switched data.
Experience with ensemble labeling, pseudo-labeling, or LLM-assisted annotation.
Familiarity with data provenance, PII/GDPR-aware pipelines, or model-improvement compliance.
Experience building custom or fine-tuned models for specific customers or domains.
Comfort working directly with research and engineering teams on shared infrastructure.

Responsibilities

Understand and characterize our data, building analyses to identify languages, conditions, domains, speakers, and quality, as well as underrepresented areas.
Design and build active-learning loops to systematically determine the most impactful areas for improvement based on performance metrics.
Develop strategies for effectively leveraging human input, optimizing workflows, tooling, and model-assisted steps to maximize the value of this scarce resource.
Establish representative benchmarking by creating curated datasets and methodologies for accurate model quality assessment across diverse real-world speech scenarios.
Challenge existing data strategies through experimentation to differentiate effective approaches from assumptions.
Ensure data consistency, cleanliness, and organization, making data accessible and understandable to non-technical teams.
Automate and standardize model adaptation processes, transforming one-off, customer-specific work into repeatable, documented pipelines.