Research Staff, Voice AI Foundations - Remote

Deepgram is a leader in Voice AI technology, providing APIs for speech-to-text, text-to-speech, and voice agent development. The company serves a large developer community and major organizations, offering cloud-based and on-premises solutions with high accuracy and low latency. Their foundation models are designed to process vast amounts of audio data efficiently, supporting innovative voice applications worldwide.

The Research Staff role focuses on pioneering new models for voice AI, specifically Latent Space Models that address data scarcity, scalability, and cost challenges. The researcher will develop neural codecs, generative models, and embedding systems to enable high-fidelity, diverse, and controllable speech synthesis, as well as synthetic data generation at scale. The work involves designing architectures and algorithms optimized for hardware and real-time inference.

Candidates should have a strong background in statistical learning, foundation model architectures, and multimodal learning. Practical experience in building large data pipelines, validating innovations through experiments, and deploying models efficiently is essential. The role requires a creative, experimental mindset, comfortable with rapid change and AI-driven workflows, and is suitable for those eager to solve complex, unsolved problems in voice AI.