Senior Data Scientist, Audio

COMPANY OVERVIEW Deepgram is bringing the world’s most natural interface, voice, to the technology we use every day. Our speech models and real-time infrastructure power AI systems that keep up with the way people actually speak. More than 5,000 organizations run conversational voice AI on Deepgram in production, including Twilio, Cloudflare, Sierra, Decagon, Vapi, Cresta, Granola, and Jack in the Box. The Deepgram Voice Agent Platform brings together speech-to-text, text-to-speech and speech-to-speech, with voice-native foundation models available through cloud APIs or self-hosted and on-premises deployments. Built for accuracy, low latency and cost efficiency, Deepgram makes high-quality voice experiences practical to build and run at scale. More than 650,000 developers have used Deepgram to process more than 200,000 years of audio. Every day, more of the world’s conversations with AI run on our models. Operating Principles: Deepgram is a research company that productizes as fast as possible. Four Operating Principles shape how we think, decide, and work together: 1. Compress before you communicate. You think it through first, then lead with the bottom line. 2. Resist entropy. You question what already exists before adding more process or tools. 3. Solve the category. You fix recurring problems at the source, not one case at a time. 4. Work at the frontier. You go after problems no one has solved yet. We look for people who already work this way, and these principles shape how we grow and evaluate our team. AI-FIRST BY DEFAULT We build AI that millions of people talk to, and we run the company the same way. AI is how we centralize work here, and is core to how we operate. That means more than using the tools. We expect you to: - Build, not just use. Design your own workflows and agents, and use AI to remove whole categories of work instead of doing the same work faster. - Own the output. AI can speed up the work. It doesn’t make anyone less accountable for it. You understand, verify, and stand behind everything you share, including what an agent produces for you. - Spend the time you get back at the frontier. The capacity AI creates goes toward problems no one has solved yet. - Keep pace. Models and tools change every month, and so will your work. You adopt what’s new quickly and drop what no longer earns its place. We judge AI by what it helps you deliver and whether you stand behind the result, not by how often you use it. In interviews, expect to walk us through a workflow you built and where you drew the line between AI and your own judgment. Deepgram is looking for a Data Scientist to sit at the intersection between research and data: people who think deeply about what conversational data is actually composed of, what makes data valuable, and how to best leverage that. Conversational audio presents incredibly rich scientific, engineering, and infrastructure challenges that are orders of magnitude harder than working with text. Speech carries speakers, accents, dialects, emotion, overlapping talk, code-switching, domain vocabulary, and acoustic conditions that range from a quiet studio to a drive-through — all with rich context that moves the conversation. In this role you'll collaborate closely with our research, engineering, and data teams to answer those questions rigorously and at scale. You'll care deeply about what makes conversational data difficult, you'll characterize it, and you'll shape the strategies that turn it into model gains. This role rewards conviction, creativity in approach, and a real appetite for overturning your own assumptions when the data says otherwise. This is a hands-on, high-leverage role with unusual latitude to define an area from first principles. It reports to the VP of Data Operations. We're looking for people who: - See "we've always done it this way" as a starting position to argue with, not a constraint to work around - Can find the one experiment that settles a question in days rather than months - Are creative about where signal hides — in metadata, in model confidence, in the failures - Have the vision to take a scrappy proof-of-concept and scale it 100x - Are obsessed with using AI to automate and amplify their own impact WHAT YOU'LL OWN - Understand and characterize our data. Build the analysis that tells us what we actually have — languages, conditions, domains, speakers, quality — and what's underrepresented. - Design and build active-learning loops. Decide what's worth working on next based on where it will move performance, and make that decision systematic rather than intuitive. - Think deeply about how to best leverage humans in the loop. Human attention is the scarcest input in this system. Design the workflows, tooling, and model-assisted steps that make it count. - Make representative benchmarking possible. Build the curated datasets and methodology that let us make honest claims about model quality across the full diversity of real-world speech.Senior - Change our minds about what data strategies actually work. Run the experiments that separate what works from what everyone assumes works. - Care about data consistency, cleanliness, and organization — and about making data legible and accessible to non-technical teams, not just to the people who built the pipelines. - Bring method and automation to model adaptation. Turn one-off, domain- and customer-specific model work into repeatable, documented pipelines. WHAT WE'RE LOOKING FOR - Hands-on work on real data pipelines and model-facing problems in data science, ML, or applied research - Strong Python and data tooling; comfort building analysis, scoring, and automation yourself - 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 — you can reason about model output quality, confidence, and error modes - A track record of turning ambiguous, messy data situations into measurable model or product improvements - You build systems others run without you in the room — a reusable harness, not a one-off notebook - Strong communication skills, especially translating complex findings for audiences who don't share your background - An active AI-tool user. Not aspirational — tell us what you use and what you've built with it It would be great if you also had: - 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. Backed by prominent investors including Y Combinator, Madrona, Tiger Global, Wing VC and NVIDIA, Deepgram has raised over $215M in total funding. If you're looking to work on cutting-edge technology and make a significant impact in the AI industry, we'd love to hear from you! Deepgram is an equal opportunity employer. We want all voices and perspectives represented in our workforce. We are a curious bunch focused on collaboration and doing the right thing. We put our customers first, grow together and move quickly. We do not discriminate on the basis of race, religion, color, national origin, gender, sexual orientation, gender identity or expression, age, marital status, veteran status, disability status, pregnancy, parental status, genetic information, political affiliation, or any other status protected by the laws or regulations in the locations where we operate. We are happy to provide accommodations for applicants who need them. Notice: We're aware of individuals impersonating Deepgram recruiters. All legitimate Deepgram recruiting communication comes from an @deepgram.com http://deepgram.com email address. If you've received a message claiming to be Deepgram, please forward it to careers@deepgram.com.