Director, Text-to-Speech Synthesis Research

This is a fully remote position, open to applicants in California, +1 more state.

📋 Description

• Take ownership of the TTS research and model roadmap.

• Determine which technical strategies can significantly enhance speech-generation quality.

• Propel advancements in neural audio modeling, prosody, expressiveness, controllability, multilingual speech, voice identity and consistency, data and training strategies, as well as post-training and inference performance.

• Ensure that research developments lead to quantifiable improvements in production.

• Evaluate research, question assumptions, design experiments, troubleshoot model failures, and address high-impact technical challenges.

• Develop evaluation and benchmarking processes utilizing automated metrics and human perceptual assessments.

• Supervise individual contributors and technical lead managers.

• Recruit and nurture researchers and research leaders while upholding a high technical standard.

• Mentor senior researchers to become technical leaders and define direction across sub-teams.

• Collaborate with engineering and product leadership to ensure ship-readiness.

• Represent Deepgram’s TTS research both internally and externally.

⛳️ Requirements

• Extensive expertise in contemporary TTS, speech generation, or audio generative modeling.

• Proven history of personally training and enhancing large-scale neural models.

• Proficient understanding of the modern speech-generation stack and the existing challenges regarding naturalness, expressiveness, controllability, robustness, voice consistency, and inference costs.

• Experience in setting research direction amidst uncertainty, prioritizing experiments, allocating compute resources and researcher time, and discontinuing ineffective methods.

• Proven experience in leading researchers and research engineers through other technical leaders, developing tech lead managers or their equivalents, and defining direction across sub-teams.

• AI as the fundamental mode of operation, with experience restructuring workflows around AI and a clear perspective on its limitations in speech research.

• Capability to articulate complex technical trade-offs to product, engineering, and executive audiences.

• Scholarly publications or academic literature related to speech processing, STT/TTS, or similar fields.

• Technical proficiency with technologies, systems, or tools used for training AI models, including PyTorch or other relevant libraries.

• Legally authorized to work in the United States; visa sponsorship requirements must be disclosed.

• No specific educational credentials required.

🏝️ Benefits

• Equity.

• Bonus.

• Base salary compensation.

• AI Notetaker interview recording/transcription, with the option to opt out without affecting candidacy.