About this role
Deepgram builds real-time APIs for speech-to-text, text-to-speech, and production-grade voice agents at scale. The company is seeking a Staff Conversational Designer to own how voice agents converse, focusing on persona, turn-taking, repair, confirmation, and pacing.
In this capacity, the person in this role will work directly with machine learning and engineering teams on tradeoffs that make voice behaviors real, including endpointing, barge-in, and latency budgets. Responsibilities also include building evaluations to measure conversational quality and publishing guidance that shows developers how to build natural interactions on the Voice Agent API. A strong candidate will have deep experience designing conversational behavior for LLM-based voice agents and assistants, alongside a solid understanding of the speech pipeline from ASR to TTS.
What you will do
- Define and maintain the persona and voice system for Deepgram voice agents to ensure coherence across various experiences and use cases.
- Collaborate with machine learning and engineering teams to design turn-taking, barge-in, and end-of-turn behavior while balancing responsiveness.
- Develop conversational repair strategies, handle no-match and no-input scenarios, and implement guardrails for high-stakes actions.
- Manage latency-aware pacing and perceived responsiveness through brevity, backchanneling, and filler speech against real-time budgets.
- Establish conversational-quality evaluations and a transcript review practice to convert production failures into a repeatable design loop.
- Construct reference agent experiences and developer-facing design guidance that demonstrate best-practice conversation on the API.
- Partner with research teams on automatic speech recognition and text-to-speech behavior to define the criteria of a good conversation.
What you need
- Bring deep experience designing conversational behavior specifically for LLM-based voice agents or assistants rather than traditional scripted IVR flows.
- Demonstrate real fluency with the speech pipeline from automatic speech recognition through large language models to text-to-speech.
- Show a proven track record of designing turn-quality, interruption, repair, and confirmation patterns that have shipped and operated successfully in production.
- Provide evidence of building quality measurement into practice through evaluations, transcript reviews, benchmarks, or structured failure-analysis loops.
- Possess exceptional writing craft encompassing sample dialogs, design guidance, and documentation that technical users can build against.
- Display experience influencing engineering and machine learning partners on behavior without having direct authority over them.
Nice to have
- Time spent on a named assistant or a production voice agent platform
- Practice with Wizard of Oz testing and sample dialog methods
- Hands-on work with evaluation tooling for large language models or voice quality
- Experience in high-stakes or regulated conversation domains where confirmation and recovery carry real cost
- Multilingual or cross-locale conversation design experience
Benefits
- Offers Equity
- Offers Bonus
Skills and tech stack
Questions candidates ask
Where is this role located and what is the remote policy?
This is a remote role based in Pacific Time, with a preference for candidates located in San Francisco.
What is the compensation range for this position?
The salary ranges from $180K to $240K, alongside offered equity and bonuses.
What kind of team will I be working with?
You will report directly to the Director of Product Design as a Staff individual contributor, working closely with ML and engineering teams.
What are the initial priorities for this position?
Your first chapter involves establishing a documented persona and voice system, alongside turn-taking and repair behavior designed and shipped with engineering.