Research Staff, Voice AI Foundations - Remote

The role

You will pioneer the development of Latent Space Models to solve data, scale, and cost challenges in voice AI. This involves building neural audio codecs, steerable generative models, and embedding systems for speech-to-speech systems.

What you will do

• Build next-generation neural audio codecs
• Pioneer steerable generative models for human speech synthesis
• Develop embedding systems to factorize codec latent space
• Leverage latent recombination to generate synthetic audio data
• Train multimodal speech-to-speech systems
• Design model architectures, training schemes, and inference algorithms

Required qualifications

• Strong mathematical foundation in statistical learning theory
• Deep expertise in foundation model architectures
• Ability to bridge theory and practice
• Ability to build data pipelines for massive datasets
• Track record of designing controlled experiments
• Experience optimizing models for real-world deployment
• History of open-source contributions or research publications in speech/language AI

Nice to have

• Mastery of Core Concepts
• Algorithmic Innovation & Implementation
• Data-Driven & Scalable Systems
• Hardware & Systems Understanding
• Rigorous Experimental Design

About the company

Deepgram is a research company that productizes as fast as possible. Their speech models and real-time infrastructure power AI systems for more than 5,000 organizations.