TTS Research Director — From Models to Production

Deepgram; Myrtle Point (OR); On-site; USD 180,000 - 260,000; Full time

Job summary

Deepgram, a leader in AI-powered speech tech, seeks a Director of Research to own the Text-to-Speech program end to end — strategy, models, and production deployment. This hands-on leadership role shapes architectures, experiments, training strategy, and team building to advance speech generation.

You will build and guide a team of researchers and engineers, aligning research bets with product needs and production constraints while maintaining a high technical bar.

Qualifications

• Deep expertise in modern TTS or audio generative modeling.
• Command of the modern speech-generation stack and related challenges.
• History of setting research direction under uncertainty.
• Experience leading researchers and research engineers through other leaders.
• AI as a default mode of work and ability to integrate it into workflows.
• Ability to explain complex tradeoffs to product and exec audiences.

Responsibilities

• Own the TTS research and model roadmap, deciding strategic directions and key bets.
• Drive advances across neural audio modeling, prosody, expressiveness, and inference performance.
• Review research, design experiments, diagnose failures, and tackle high-leverage problems.
• Build evaluation benchmarks that explain improvements with metrics and perceptual tests.
• Lead a team of researchers and tech-lead managers, setting direction across sub-teams.
• Partner with engineering and product leadership on ship-readiness and external representation.

Job description

Deepgram, a leader in AI-powered speech tech, seeks a Director of Research to own the Text-to-Speech program end to end — strategy, models, and production deployment. This hands-on leadership role shapes architectures, experiments, training strategy, and team building to advance speech generation.

You will build and guide a team of researchers and engineers, aligning research bets with product needs and production constraints while maintaining a high technical bar.