Machine Learning Scientist

This is a fully remote position, open to applicants in United States.

📋 Description

• Create, train, and assess speech synthesis models, both autoregressive and non-autoregressive.

• Initiate research into full-duplex and half-duplex multi-modal architectures, including unified S2S systems.

• Select and refine speech representations: neural codecs, semantic tokens, mel features, and continuous latents.

• Establish rigorous evaluations, both objective and perceptual, maintaining high standards for quality and prosodic control.

• Partner with our linguists on TTS frontend behavior to ensure that modeling and frontend choices work in harmony.

⛳️ Requirements

• Extensive knowledge of the speech synthesis literature, both contemporary and historical — including Tacotron, FastSpeech, VITS, VALL-E, and the codec-LM lineage. Possess insights on what has been effective and why.

• Practical experience with neural codecs (EnCodec, DAC, Mimi, etc.) and various representation options.

• Proficient in full- or half-duplex multi-modal modeling (Moshi, LLaMA-Omni, streaming S2S).

• Exceptional attention to detail regarding data quality, able to identify when an annotation pipeline is deteriorating or when an evaluation set is compromised.

• Eager to engage in hands-on data and training tasks, with the initiative to develop pipelines to alleviate manual processes for the team.

• Familiarity with TTS frontend components (G2P, normalization, prosody) and experience collaborating with linguists.

• Strong foundational knowledge of PyTorch, comfortable with training loops, distributed training, and model internals.

• PhD or equivalent research experience in speech, audio, ML, or computational linguistics, or a proven track record that renders formal credentials unnecessary.

• Experience in multilingual TTS.

• Background in prosody or paralinguistics.

• Published research in speech, audio, or core ML venues.

• Experience transitioning research models to production, including quantization, distillation, and streaming inference.

🏝️ Benefits

• Competitive base salary plus significant early-stage equity.

• Remote-friendly work environment.

• Visa sponsorship available.

• Access to a proprietary, full-duplex, studio-quality conversational speech corpus.

• Provision of computing resources and tools necessary for the work.

• Opportunity to have a direct impact on the future of voice AI.