Conduct foundational research on latent space models for voice AI, including neural audio codecs, generative speech, interpretable embeddings, synthetic data generation, multimodal speech-to-speech systems, and hardware-efficient model architectures. Design mathematical formulations, scalable data pipelines, controlled experiments, and real-time inference solutions for large-scale audio datasets.

Leads Deepgram’s end-to-end Text-to-Speech research program, owning research strategy, technical direction, model development, evaluation, and production deployment. Advances neural audio modeling, prosody, expressiveness, multilingual generation, voice consistency, controllability, and inference performance. Builds and develops research teams, manages technical leaders, prioritizes experiments and compute, evaluates model quality, and partners with engineering and product leadership to ship production-grade models.